Download Munshi_washington_0250E_12611

Survey
yes no Was this document useful for you?
   Thank you for your participation!

* Your assessment is very important for improving the work of artificial intelligence, which forms the content of this project

Document related concepts

Astrophysical X-ray source wikipedia, lookup

White dwarf wikipedia, lookup

Nucleosynthesis wikipedia, lookup

Planetary nebula wikipedia, lookup

Weak gravitational lensing wikipedia, lookup

Gravitational lens wikipedia, lookup

Standard solar model wikipedia, lookup

Cosmic distance ladder wikipedia, lookup

Accretion disk wikipedia, lookup

P-nuclei wikipedia, lookup

Hayashi track wikipedia, lookup

Main sequence wikipedia, lookup

Cygnus X-1 wikipedia, lookup

Stellar evolution wikipedia, lookup

H II region wikipedia, lookup

Astronomical spectroscopy wikipedia, lookup

High-velocity cloud wikipedia, lookup

Star formation wikipedia, lookup

Transcript
c
°Copyright
2013
Ferah Munshi
Star Formation
in N-Body + SPH Simulations
Ferah Munshi
A dissertation
submitted in partial fulfillment of the
requirements for the degree of
Doctor of Philosophy
University of Washington
2013
Reading Committee:
Thomas Quinn, Chair
Fabio Governato
Alyson Brooks
Program Authorized to Offer Degree:
Astronomy
University of Washington
Abstract
Star Formation
in N-Body + SPH Simulations
Ferah Munshi
Chair of the Supervisory Committee:
Professor Thomas Quinn
Astronomy
The primary focus of my thesis work is to study star formation using a series of high resolution cosmological N-body Simulations. Specifically, I have studied the total stellar-to-halo
mass ratio as a function of halo mass for a new sample of simulated field galaxies using fully
cosmological, LCDM, high resolution SPH + N-Body simulations carried to the present
time. I find there is extremely good agreement between the simulations and predictions
from the statistical Halo Occupation Distribution model presented in Moster et al. (2012).
This is due to a combination of systematic factors: a) gas outflows that reduce the overall
SF efficiency and b) estimating the stellar masses of simulated galaxies using artificial observations and photometric techniques similar to those used in observations. My analysis
suggests that stellar mass estimates based on photometric magnitudes underestimate the
contribution of old stellar populations to the total stellar mass, leading to stellar mass errors
of up to 50% for individual galaxies and highlight the importance of using proper techniques
to compare simulations.
Additionally, my work examines the pressure of the star-forming interstellar medium
(ISM) of simulated high-resolution Milky-Way sized disk galaxies, using a kinematic decomposition of these galaxies into present-day bulge and disk components. I find that the
typical pressure of the star-forming ISM in the present-day bulge is higher than that in the
present-day disk by an order of magnitude. Additionally, the pressure of the star-forming
ISM in the early protogalaxy is on average, higher than ISM pressures after z = 2. This
explains the why the bulge forms at higher pressures: the disk assembles at lower redshift,
when the ISM is generally lower pressure and the bulge forms at higher redshift when the
ISM is at higher pressures. If ISM pressure and IMF variation are tied together as suggested
in studies like Conroy van Dokkum (2012), these results could indicate a time-dependent
IMF in Milky-Way like systems.
Finally, my thesis work addresses the question of how well observational star formation
indicators measure the true underlying star formation in dwarf galaxies. In particular, I
examine Hα and the UV continuum as star formation indicators and study the timescales
upon which these indicators accurately measure the underlying star formation of the galaxy.
Additionally, I examine the effects of star formation prescription and resolution on applying
observational indicators to simulations.
TABLE OF CONTENTS
Page
List of Figures . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . iii
List of Tables . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . viii
Chapter 1:
Introduction . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . .
1
1.1
Background Information . . . . . . . . . . . . . . . . . . . . . . . . . . . . . .
1
1.2
Outline . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . 17
Chapter 2:
Reproducing the Stellar Mass/Halo Mass Relation in Simulated ΛCDM
Galaxies: Theory vs Observational Estimates . . . . . . . . . . . . . . 19
2.1
Introduction . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . 20
2.2
The Simulations . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . 24
2.3
The Stellar Mass to Halo Mass Relationship . . . . . . . . . . . . . . . . . . . 29
2.4
Conclusions . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . 33
2.5
Supplemental Analysis . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . 34
Chapter 3:
The Pressure of the Star Forming ISM in Cosmological Simulations . . 48
3.1
Introduction . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . 48
3.2
The Simulations and Analysis . . . . . . . . . . . . . . . . . . . . . . . . . . . 50
3.3
Results . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . 53
3.4
Summary . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . 55
3.5
Supplementary Analysis . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . 56
Chapter 4:
Dwarfs and Star Formation (SF) Indicators: Are SF Indicators Sensitive to Timescale? . . . . . . . . . . . . . . . . . . . . . . . . . . . . . 67
4.1
Introduction . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . 67
4.2
The Simulations and Analysis . . . . . . . . . . . . . . . . . . . . . . . . . . . 68
4.3
Results . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . 70
i
4.4
Summary . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . 75
Chapter 5:
Conclusions and Future Work . . . . . . . . . . . . . . . . . . . . . . . 80
5.1 Conclusions . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . 80
5.2 Future Work . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . 82
Bibliography . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . 88
ii
LIST OF FIGURES
Figure Number
1.1
Page
Plot of the different functional forms of the IMF used in the literature.
Figure courtesy of Ivan Baldry. . . . . . . . . . . . . . . . . . . . . . . . . .
7
1.2
Plot of the mass-metallicity relationship in simulations, compared with observations. This plot shows the agreement between simulations (data points)
and observations (solid line). Plot from Brooks et al. (2007). . . . . . . . . 15
1.3
Plot of cusp vs. core in collisionless simulations, compared with simulations
that include baryons. The black dot-dashed line shows the cuspy profile
that represents the overproduction of low angular momentum DM. The
inclusion of baryons and the effect of feedback shows that simulations can
now reproduced cored profiles that are often observed. Plot from Governato
et al. (2012). . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . 16
2.1
Baryonic fraction with respect to the cosmic ratio, for simulated field galaxies as a function of stellar mass, measured at z=0. Circles are the “direct
from simulation” results, including all gas and stars within R200 . Triangles are the “observable” baryon fractions, includind all stars and all the
‘observable’ cold gas (defined as 1.4 × (HI+H2), within R200 ). The empty
symbols are galaxies with no observable gas (cold gas mass < 100 M⊙ ).
Symbol sizes represent the different mass resolutions of galaxies in the sample. Smaller symbols are the higher mass resolution by a factor of 8 when
compared to the larger symbols. Galaxies below 108 M⊙ lose a significant
fraction of baryons due to heating from the cosmic UV background and SN
feedback. . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . 37
2.2
The stellar mass ratio between the galaxies simulated with the old ‘low density SF threshold’ and the new sample. In the new sample SF is regulated
by the local abundance of molecular hydrogen, resulting in feedback significantly lowering the total SF efficiency. All quantities as measured directly
from the simulations. . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . 38
iii
2.3
The cold gas mass as a function of stellar mass. Simulations vs. SHIELD
and ALFALFA data. The HI mass of each galaxy in the simulated sample is
plotted vs the SDSS r-band magnitude and compared to two samples from
nearby surveys. Red solid dots: simulations. Diamonds: ALFALFA survey.
Asterisks: SHIELD survey. While feedback removes a large fraction of the
primordial baryons, the simulated galaxies have a high gas/stellar mass
ratio, comparable to the observed samples. Most of the cold gas resides
within a few disk scale lengths from the simulated galaxy centers. Figure
courtesy of A. Brooks. . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . 39
2.4
The Stellar Mass vs Halo Mass. Black Solid Dots: The SHM relation from
our simulations set with stellar masses measured using Petrosian magnitudes and halo masses from DM-only runs. This procedure mimics the
one followed in M12. Open Dots: Unbiased stellar masses measured directly from the simulations. Solid Line: Observational results from M12.
Symbol sizes represent the different mass resolutions of galaxies in the sample. Smaller symbols are the better mass resolution by a factor of 8 when
compared to the larger symbols. . . . . . . . . . . . . . . . . . . . . . . . . 40
2.5
Top Panel: Halo mass ratio of galaxies in runs with baryons and SF vs
DM-only runs. Individual halos in DM–only runs are typically 30% more
massive than their counterparts in simulations with gas physics and SF. The
effect is smaller in more massive halos, where baryon loss due to feedback
is less (see also Sawala et al. (2012)). The dashed horizontal line marks the
ratio if halos had a 100% baryon loss. Middle Panel: Estimated vs. True
Stellar Mass as a function of halo mass. The stellar mass using artificial
Petrosian magnitudes and measured using the photometric method in (Bell
& de Jong, 2001) vs the “true” Stellar mass measured directly from the
simulations. Stellar masses measured using the photometric method in
(Bell & de Jong, 2001) in combination with the flux loss from applying the
petrosian magnitudes are underestimated by about 50% across the range of
galaxy masses in our study. . . . . . . . . . . . . . . . . . . . . . . . . . . . 41
2.6
Stellar mass within Petrosian radius using colors vs straight from simulations. This highlights the contribution of using observational techniques in
the underestimation of stellar mass, after the Petrosian radius is applied.
The use of a fixed aperture underestimates mass by a further 10-20%, as
previously estimated in Blanton et al. (2001) . . . . . . . . . . . . . . . . . 42
iv
2.7
The Stellar Mass vs Halo Mass. Black Solid Dots: The SHM relation from
our simulations set with stellar masses measured using Petrosian magnitudes and halo masses from DM-only runs. This procedure mimics the one
followed in M12. Open Dots: Unbiased stellar masses measured directly
from the simulations. Solid Green Line: Observational results from M12.
Solid Red Line: Results from Behroozi 2012. . . . . . . . . . . . . . . . . . . 43
2.8
Comparison of dark matter halo mass between a baryonic run and a collisionless run. This figure shows that in the larger halos, the change in virial
mass can be accounted for just by the change in overdensity after baryons
are lost. However, in smaller halos, this is not sufficient: these halos must
also suffer from reduced accretion, thus the scatter below 1. Figure courtesy
of Alyson Brooks. . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . 44
2.9
Comparison of dark matter halo virial radius between a baryonic run and
a collisionless run. This figure, like the previous, shows that the change
in virial radius is, in larger halos, due to the change in overdensity due to
baryonic physics. In the smaller halos, this is not the case and these halos
must have reduced accretion to account for the change in virial radius.
Figure courtesy of Alyson Brooks. . . . . . . . . . . . . . . . . . . . . . . . 45
2.10
Anyalysis of mass recovered as a function of stellar population age when
using M/L ratios to estimate stellar mass. In this figure, we use B-V color,
as in the analysis for the SMHM relation. The first point is so high because
the stellar population is so old that the B-V colors are undefined. The
x-axis is time in simulation units, ranging from 0 to 13.7 Gyrs. . . . . . . . 46
2.11
Stellar mass in a given halo mass. This figure shows how the star formation
efficiency increases toward milky way halo masses. . . . . . . . . . . . . . . 47
3.1
Top Panel: Distribution of pressures for bulge and disk in one of the simulated galaxies, h986. Note that the peak of the distribution of pressures
of the bulge is higher than the peak of the pressure distribution of disk
stars. Middle Panel: SFHs for the bulges of the 3 galaxies not shown in
this manuscript. Note that like h986 shown in Figure 3, the bulges form
early in the galaxy’s history. Bottom Panel: SFHs for the disks of the 3
galaxies not shown in this manuscript. As with the bulges, these galaxies
also follow the same trend as h986, with disk star formation occuring later
in the galaxy’s history. . . . . . . . . . . . . . . . . . . . . . . . . . . . . . 58
v
3.2
Phase Diagrams for bulge (top) and disk (bottom) during a star formation
event, color coded by pressure. Star formation events for each component
were selected based on contribution to each components’ overall growth.
Hotter colors are higher pressures, cooler colors are lower pressures. Note
that high pressures are driven by high densities in the bulge. In the bottom
panel we show the metallicity of the gas that formed both the bulge and
disk stars versus its density at the time of star formation (note that both
density and metallicity are smoothed over hundreds of parsecs). . . . . . . 59
3.3
Top Panel: Star formation rates for each of the dynamical components
of h986. Bottom Panel:Pressure vs. Time for each of the components.
This highlights the redshift dependence of the pressure of the star forming
ISM: early on, stars are forming at higher pressures, regardless of which
component they belong to at z = 0. Note also the peaks in pressure are
present when the SFH is peaking, in bulge stars. The big bursts in the
bulge SFH correspond to major mergers in the galaxy’s history. . . . . . . . 60
3.4
Top Panel: Pressure vs. formation radius for the bulge and disk, over the
galaxy’s whole history. Bottom Panel:Pressure vs.formation radius for each
of the components, for stars that formed before 6 Gyrs (when the pressure
of the ISM was higher for both components. This figure highlights that
formation radius is not the underlying cause of the pressure differential
between bulge and disk and that on average, bulge stars are forming at
higher pressures than disk stars. The bottom panel shows that in the first
half of the galaxy’s history, stars are forming at higher pressures in general,
regardless of component and formation radius. . . . . . . . . . . . . . . . . 61
3.5
Distribution of metallicities for bulge and disk in one of the simulated galaxies. Note that the peak of the distribution of pressures of the bulge and
disk are nearly the same. . . . . . . . . . . . . . . . . . . . . . . . . . . . . 62
3.6
Distribution of H2 mass fractions for bulge and disk in one of the simulated
galaxies. Note that the peak of the distribution of pressures of the bulge
and disk are nearly the same. . . . . . . . . . . . . . . . . . . . . . . . . . . 63
3.7
Average H2 mass fractions for the bulge as a function of time, during a star
formation event. . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . 64
3.8
Average H2 mass fractions for the disk as a function of time, during a star
formation event. . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . 65
3.9
Phase diagram for bulge stars color coded by H2 fraction. Blue points are
high H2 fractions, and red points are low H2 fractions. . . . . . . . . . . . . 66
vi
4.1
4.2
4.3
4.4
4.5
4.6
4.7
4.8
Simulated observation of Hα emission from SUNRSISE SED. The lineflux
from this line is used to derive SFR in simulated dwarfs, normalized by
continuum measurements in the simulated SED. . . . . . . . . . . . . . .
Simulated observation of 24 micron emission from SUNRISE. The flux from
thses images is used to derive SFR in simulated dwarfs. . . . . . . . . . .
Simulated observation of UV emission from SUNRSISE generated image,
including dust reprocessing. The flux from such images are utilized to derive
SFRs. . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . .
Ratio of SFRs derived from the UV continuum and from Hα. Stellar mass
is on the x-axis, and the ratio of SFRs from the various indicators are on
the y-axis. Green squares are the simulations, and the black points are the
data from Weisz et al. (2012). . . . . . . . . . . . . . . . . . . . . . . . .
Ratio of SFRs derived from the UV continuum and from Hα, utilizing lower
resolution simulations, with only metal-line cooling. Each color represents
a different dwarf simulation; if more that one point exists for a given color,
one point is z=0 and the other is at z=0.5. The greyed out points in the
backgroud once again are the observational data from Weisz et al. (2012).
Top Panel: Star formation rate for the lower resolution, metal-line cooling
run. Bottom Panel:Star formation rate for the high resolution, H2 star
formation run. . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . .
Measured Hα SFRs versus actual SFRs integrated over various timescales.
Each color represents a different halo. Hα SFRs measured using the lineflux from the simulated SED. The measured SFRs come directly from the
simulations, integrated over the timescale shown on the x-axis. This plot,
although inconclusive, shows that the SFR does change as a function of
timescale it’s integrated over, although no single timescale appears to match
with the Hα SFRs in this mass range. . . . . . . . . . . . . . . . . . . . .
Measured UV continuum SFRs versus actual SFRs integrated over various timescales. Each color represents a different halo. Hα SFRs measured
using the lineflux from the simulated SED. The measured SFRs come directly from the simulations, integrated over the timescale shown on the
x-axis. This plot, although inconclusive, shows that the SFR does change
as a function of timescale it’s integrated over, although no single timescale
appears to match with the UV SFRs in this mass range. . . . . . . . . .
vii
. 71
. 72
. 73
. 74
. 75
. 77
. 78
. 79
LIST OF TABLES
Table Number
3.1
Page
Description of simulations utilized in this analysis. . . . . . . . . . . . . . . . 52
viii
ACKNOWLEDGMENTS
To the Beard, Professor Thomas Quinn, who made me the scientist I am today. To
Professor Fabio Governato, for giving the pushes and nudges that I needed to take this
thesis to completion, in spite of all the complaining I did along the way. To the rest of
my committee: Professors Julianne Dalcanton and Alyson Brooks, the strongest and most
successful women that I am lucky to have as mentors.
I could imagine no other place to have completed graduate school than the University
of Washington. I admire, respect and have grown very fond of all the peers that I have met
through this program: they have grown to be like family and I will miss them a lot after I
leave. Most notably, I will miss all the officemates I have had along the way especially those
I have now, Michael Tremmel and Kristen Garofoli. And the best roommate I’ve ever had:
Vaishali Bhardwaj, who went from my frenemy at Berkeley, to one of my best friends, and
bridesmaid at my wedding.
It has been more than amazing to live in city like Seattle. I’ve enjoyed the food, the
outdoors and found a hobby that has brought me great joy: crossfit. A giant shoutout to
the wonderful and thoughtful folks I have met at the Lab. I will miss doing really hard
things and throwing heavy weights around with you guys.
Last but not least, is my family. Without John and Ivan, I would not have had the
motivation and support necessary to get here.
ix
DEDICATION
To my loving husband, John
and the cutest pug in the world, Ivan
for making me feel safe and loved.
x
1
Chapter 1
INTRODUCTION
Galaxy simulations enable astronomers to track the numerous nonlinear processes that
lead to a present-day galaxies. N-Body simulations provide explicit spatially resolved description of the physical processes involved in galaxy formation with a minimum set of
assumptions. As the internal structure of galaxies holds precious clues on their assembly
and subsequent evolution, the use of N-body simulations is then mandatory. Only recently
have galaxy simulations been able to achieve sufficient dynamical range and physical detail
to resolve the internal structure of individual galaxies. Dark matter (DM) simulations have
informed our understanding of the build-up of structure in our present-day universe but in
order to compare to observations, we must consider simulations that contain baryons: stars,
the various types and phases of gas and their subsequent interactions, including feedback.
Specifically, we need to understand how, when and where stars are forming from gas, and
the effect that the evolution of stars has upon the gas content of our simulated galaxies. In
this thesis I specifically explore star formation in cosmological simulations, including: how
galaxies populate their dark matter halos, how the interstellar medium (ISM) structure
and star formation are related in disk galaxies and observational biases in measuring star
formation rates in observed galaxies.
1.1
1.1.1
Background Information
Hierarchical Galaxy Formation and Growth
On the largest scales, galaxy formation is determined by the accretion of matter into potential wells. The favored model for this process is hierarchical formation within the Λ-
2
Cold-Dark-Matter (LCDM) paradigm wherein small structures (halos) form first, in areas
of enhanced potential, and each of these halos merge together into deeper potential wells
caused by larger potential fluctuations. As time passes, structure builds from small halos
to larger halos, where the larger halos are the result of mergers and growth from smaller
halos.
The general picture of galaxy formation begins with an early universe relatively evenly
distributed with dark and baryonic matter. Tidal torquing spins up both dark and baryonic
matter into solid body rotation which both begin to collapse under the influence of gravity
in overdense regions. The collapse of dissipationless dark matter is stopped by virialization;
baryonic matter however, continues to collapse and decouples from the dark matter, settling into a rotating disk. This simple top-down picture is favored by many analytic models
including White & Rees (1978); Fall & Efstathiou (1980); Dalcanton (1998) and can explain
a variety of features of galaxies, including the Tully-Fisher relation, the Luminosity-Size relation and the mass-metalicity relation. However, these models make the basic assumptions
of smooth accretion and no mergers, which is an incomplete picture.
Mergers and interactions between galaxies are an essential ingredient of galaxy formation
and evolution: Toomre & Toomre (1972) was among the first to recognize that mergers can
drive the evolution of galaxy types and subsequent simulations that include gas dynamics,
star formation and feedback have further demonstrated that both major and minor mergers
can affect the the long-term survivability of disks, bulge growth, and structural evolution
of galaxies (Springel et al., 2005a).
The rate of gas cooling into halos determines its accretion into dark matter halos and its
subsequent collapse into the first generation of stars. The first halos accreted pristine gas,
unpolluted by metal enrichment from stellar evolution: 75% Hydrogen and the remaining
in Helium (no metals). Both elements initially cooled through recombination, collisional
excitation and emission, in addition to Bremsstrahlung radiation. However, at lower temperatures (less than 104 K), the dominant method of cooling is molecular hydrogen cooling.
This low temperature cooling was the dominant cooling mechanism for forming the first
3
generation of very massive stars (Pop III stars) in early DM halos. These are also the stars
responsible for the reionization of the universe.
After the formation of the first stars, galaxy halos continue to merge and grow until they
are organized into groups, clusters and field galaxies, located in filaments and surrounded by
voids. The most massive potential wells house the most massive structures: galaxy groups
and clusters contain giant elliptical galaxies in their centers, within a large potential well,
whereas spiral galaxies and dwarfs are found in the outskirts or even in isolation. These
wide variety of structures and morphological types show even greater variety when their
baryon content, colors, star formation histories and metallicities are compared: Ellipticals
tend to be gas poor, very red, and housing an old stellar population, while many spirals and
dwarfs may be quite blue, gas rich, and still actively forming stars. Given all the variety of
physics that goes into the formation and evolution of these halos (i.e., merger history, gas
accretion history, star formation), the necessity and usefulness of modeling using N-body
simulations becomes clear.
1.1.2
Simulations: Connecting Dark Matter and Baryons
The easiest and least computationally expensive simulations are Dark Matter (DM) only
(collisionless) simulations. These simulations are incredibly vital to model halo growth as
halos enter the non-linear regime. Collisionless simulations only need to model gravitational
interactions, and as stated above, are thus simple, fast and inexpensive. This type of
simulation has been used to study hierarchical formation of galaxies in large volumes in the
context of LCDM and have been key for understanding the important aspects of galaxy
formation including but not limited to: the DM halo mass function (Jenkins et al. (1998);
Springel et al. (2005a); Warren et al. (2006a); Reed et al. (2007); Warren et al. (2006b);
Boylan-Kolchin et al. (2009)), the mass power spectrum ((Davis et al., 1985; White et al.,
1987; Jenkins et al., 1998; Springel et al., 2005a; Boylan-Kolchin et al., 2009; Teyssier et al.,
2009)) and determining merger histories as a function of environment and mass (Springel
et al. (2005a); Boylan-Kolchin et al. (2009); Fakhouri et al. (2010); Hopkins et al. (2009)).
4
Collisionless simulations have also been key in understanding DM halo substructure, DM
density profiles of a given halo, and halo angular momentum as it relates to halo shape.
However, collisionless simulations are limited as it has been shown that baryons do in fact
have an affect on DM and vice versa.
Furthermore, in order to compare simulations to observations, we must model the
baryons, which are observable, to make an ’apples-to-apples’ comparison. There are many
ways in which one can compare a simulation to observations, and most require comparing
observations of stellar light to the stellar content in simulations. These methods include:
• N-Body + hydrodynamic simulations: fully self-consistent simulated models of galaxy
formation that include gravity and gas dynamics that evolve as a function of time (see
e.g., Katz (1992); Sijacki et al. (2007))
• Halo Occupation Distribution (HOD) methods & Abundance Matching methods:
match observed luminosities and galaxy distributions to collisionless simulation (see
e.g. Wang & White (2009); Guo et al. (2010); Moster et al. (2010)). These are
discussed in detail within Chapter 2.
• Semianalytic Models: use simple equations and analytic models to describe baryonic
processes for DM halos of a given mass and size (see e.g. Mo et al. (1998); Somerville
et al. (2008); Guo et al. (2010)). These usually build from the outputs of a collisionless
simulation.
Of the above methods, HOD methods and abundance matching have the advantage
of clearly matching observations by definition but do not contain information about the
systems’ evolution and the physical processes that result in the observation; rather they only
contain information about the system as it was at the time of the observation. Semianalytic
and hydrodynamic methods have the advantage of modeling the physical processes and
tracing a system’s evolution but generally require a large number of free parameters that
need to be tuned. Between SAMs and N-body + hydrodynamics simulations, the latter
5
require the fewest free input parameters (knobs) and thus can explicitly trace physical
processes and evolution with a minimum set of input assumptions.
There are two leading methods for implementing an N-body + hydrodynamics simulations: the lagrangian method (SPH) and the eulerian method (AMR). The former uses
smoothed particle hydrodymics (SPH) and uses the equations of fluid dynamics to calculate
gas temperature, pressure and density for individual particles that trace the gas. The latter
uses an adaptive mesh (AMR) in which the hydrodynamic equations are solved on a grid
of varied resolution, over discrete volumes. The simulations utilized and described in this
thesis are SPH simulations, run using the parallel Nbody+SPH code, GASOLINE.
N-body + hydrodynamics simulations also include subgrid models for other physical processes in the galaxies, not the just gas, including but not limited to: star formation, feedback
(supernovae, young stars), metal enrichment. These parameters are considered ’subgrid’ as
the physics of these processes may be limited by the resolution of the simulations (the size
of the mesh, or the softening length in an SPH simulation). However, even with resolution
limitations, hydrodynamic simulations have had vast success in reproducing observed properties of present day galaxies. For example, simulated galaxies lie on the observed scaling
relations for disk galaxies as a function of time, e.g., the size – luminosity and size – velocity
relations (Brooks et al. 2010), and from work presented in this thesis, the Stellar Mass-Halo
Mass (SMHM) relation (Munshi et al., 2013b). The galaxies also reproduce the observed
mass – metallicity relation for galaxies as a function of redshift (Brooks et al., 2007).
As stated above, the simulations in this thesis were run using GASOLINE. The following
is a description of the parameters described within the GASOLINE code:
• an SPH treatment of hydrodynamics and diffusion processes
• Scaling up to several thousands of CPUs- allowing simulations to run faster and more
efficiently.
• Star Formation with different IMFs, O and Fe metal enrichment.
• Gas heating from SN feedback, super-massive black holes (SMBHs) and a cosmic UV
background including QSOs.
6
• H2 shielding and metal line cooling; SF related to the local abundance of H2 .
1.1.3
Star Formation in the Context of the Interstellar Medium (ISM) and the Initial Mass
Function (IMF)
Star formation and stellar evolution are cyclic processes: stars are born from the gas and
dust that exists between stars, known as the interstellar medium (ISM) and during a star’s
life, depending on the star’s total mass, much of that material is returned to the ISM
through stellar winds and explosive events like Supernovae (feedback events). The ISM
contains many components: dust in the form of silicates and polycyclic aromatic hydrocarbons (PAHs), hydrogen in the form of HI gas and also hydrogen in its molecular form: H2 .
The collapse of gas into the stars begins with clouds of molecular gas (molecular clouds)
which are enoromous complexes of dust and gas where temperatures are typicaly tens of
kelvin and masses may reach upward of 106 solar masses. A cluster of star(s) begins to
collapse from such a molecular cloud when the gravitational potential energy of the cloud
dominates its internal energy (ie, it reaches it’s Jean’s length), wherein the cloud is essentially collapsing under isothermal freefall, known as homologous collapse.
λJeans = (
15kB T 0.5
)
4Gπµρ
(1.1)
In the above equation, T represents cloud temperature, ρ is density of the cloud, and µ
is the mean molecular weight. As the cloud collapses, the density increases, decreasing the
Jean’s Mass- which leads to the fragmentation of molecular clouds, forming multiple smaller
collapsing cores. Collapse halts when the cores reach sufficient surface density to become
optically thick. This causes the gas to heat up resulting in sufficient pressure support to
stabilize the young protostar- eventually the temperatures increase enough such that fusion
is sparked in the core and the protostar becomes a young stellar object and will continue it’s
evolution, based on its mass, along the H-R diagram. The initial mass function (IMF) gives
the statistical distribution of masses of stars that form in a single star formation collapse
event. Typically, the IMF is parameterized as a power law, a series of broken power laws,
7
Figure 1.1 Plot of the different functional forms of the IMF used in the literature. Figure
courtesy of Ivan Baldry.
or a log-normal distribution, in stellar mass. The IMF is crucial in interpreting the nature
of galaxies because the light we observe is dominated by the highest mass stars, while the
total mass in stars is dominated by the lower mass regime. If we know the IMF and the age,
it is then straightforward to calculate parameters such as mass from photometry. Similarly,
when looking at tracers of star formation, the IMF allows us to convert from luminosity to
star formation rate. The IMF is generally assumed to be constant and universal: it does
not vary within galaxies or between galaxies (Gilmore et al., 2000) and this has been shown
8
to be the case in star clusters (see e.g. Kroupa & Weidner (2003)) and since most star
formation occurs in star clusters, a universal IMF seems plausible (Lada & Lada (2003)).
It is important to note, however, that PopIII stars are an exception to the universality of
the IMF as ther are no observations of zero metallicity stars, indicating that all early stars
must have been high mass.
The initial mass function was first derived by Ed Salpeter in 1955, where he derived a
power law functional form, with α as 2.35, where m is stellar mass, and N is the number of
stars:
dN = m−α dM
(1.2)
Integrating this simple form of the IMF shows us that majority of stellar mass is in low
mass stars, while the luminosity is dominated by the most massive stars. One way to derive
the IMF is to use field stars. To do this, first a mass function for main sequence stars is
found. This mass function is then integrated over the vertical dimension of the disk. This
method was first utilized by Miller & Scalo (1979). While the functional form above has
much success, it is important to note that there is substantial flattening of the measured
IMF at a charateristic mass of 1Modot . Thus it is very common for the IMF to be represented
by a broken power law: one to explain the low mass end and another to explain the high
mass end, marked with a transition at the characteristic mass. Observations of regions of
active star formation show that the IMF is set very early in the star formation process.
It is unknown when exactly this mass distribution is set though there is indication that it
is possibly set at the prestellar core stage from both theory and observations (Hopkins et
al. (2013), Testi & Sargent (1998), Johnstone et al. (2000)). There are many mechanisms
suggested as to what sets this mass distribution including: gravitational fragmentation
(yields an incomplete mass spectrum), turbulence (the energy cascade with a molecular
cloud sets the mass scales for collapse), gas accretion, and feedback (newly forming stars
have a tremendous effect on their birth cloud via feedback processes). It is likely that it is
a combination of the above processes that dictates the full spectrum of masses of the IMF.
Bonnell et al. 1998 suggest that the broad peak of the characteristic mass of the IMF is
9
set by gravitational collapse, that the high mass, Salpeter like slope is set by accretion and
that at the low mass end, it is still unknown.
The origin of the stellar initial mass function (IMF) is paramount to our understanding
of star formation, stellar evolution and feedback and galaxy formation. The IMF influences
most of the observable properties of both stellar populations and galaxies. Detecting variations of the IMF will provide deep insights into the process by which stars form including
but not limited to: the origin of the stellar mass scale, the effects of metallicity and environment and the energetics of feedback. Additionally, the IMF is a key ingredient into a huge
range of models of all the above phenomena, and a necessary assumption when deriving
physical parameters from observations. Despite being such a vital ingredient, the origin
and variations of the IMF still remain poorly understood.
In particular, of critical importance, is the question of whether the IMF is universal or
whether the IMF is sensitive to the initial conditions of star formation- i.e., the structure of
the ISM in which the stars are forming (see e.g. Kroupa et al. 2011). Growing observational
evidence suggests that the high mass behavior of the IMF is uniform, including observations
of the IMF in the Magellanic Clouds (Bastian et al., 2010; Chabrier, 2003). However at the
low mass end, there are many indications, both observationally and theoretically, that there
may be a variation in the IMF. For example, Conroy & van Dokkum (2012); van Dokkum &
Conroy (2011) show that the IMF in these systems is bottom heavy using gravity sensitive
absorption lines in the cores of giant elliptical galaxies. This has also been independently
suggested by kinematic and lensing data (Treu et al., 2010; Cappellari et al., 2012; Dutton
et al., 2013). As these systems formed their stars at high redshift, these studies give us
insight into the time-evolution of the IMF. Observationally, Conroy & van Dokkum (2012)
show that the mass to light ratios of spheroidal systems indicate a more bottom heavy IMF
at higher pressures, and at higher SFRs. This indicates that ISM pressure and the intensity
of star formation are both key in understanding how and where stars form- and whether or
not the IMF is varying.
10
1.1.3.1
Connecting Star Formation in Simulations to Observations
Galaxy stellar masses and instantaneous star formation rates (SFRs) of individual galaxies
provide key information about their formation histories and comparisons betweens star
formation histories (SFHs) and SFRs of different galaxy populations can clarify relationships
among stellar populations. These quantities are provided directly by theoretical models and
specifically by simulations, which can keep into account the complex 3-D geometries of the
assembly of galaxies. However, one cannot directly observe the SFR of a galaxy; rather, it
is inferred from spectra, broadband photometry (L(Hα), L(FUV)) or from resolved stellar
populations (CMDs) (see e.g. Dalcanton et al. 2009). Various parts of the electromagnetic
spectrum have been used to measure SFRs, including: integrated light measurements in
the UV, far infrared (FIR) measurements, or nebular recombination lines. These are direct
tracers of the young stellar population. All three equations that follow relating SFR to
various components of the EM spectrum come from Kennicutt (1998), a review of all SFR
indicators.
The ultraviolet continuum is dominated by young stars, so the SFR scales directly with
luminosity. Typically, measurements are made longward of the Lyman Alpha Forest but
short enough to minimize contamination from older stellar populations. The conversion
between UV flux over a given wavelength is derived using population synthesis models which
is dependent on calibration methods, IMF assumption, star formation timescale choice and
the choice of stellar libraries. In general, it is assumed that SFR has remained constant
over time scales longer than the lifetimes of the UV emitting population, and usually adopt
a Salpeter IMF. The main advantages of using this method include that it is directly tied
to the photospheric emission of the young stellar population and can be applied to starforming galaxies over a wide range of redshifts. However, this method is highly sensitive to
extinction and the adopted IMF. Typical extinctions can be up to 3 magnitudes (Kennicutt
1998). The equation relating SFR to the UV continuum relates the luminosity, L in the
UV continuum to the SFR via a constant that is based on an assumed IMF and stellar
population model:
11
SF R(Msun yr−1 ) = 1.4 × 10−29 Lν (ergss−1 Hz −1 )
(1.3)
Star formation can also be measured from nebular emission: nebular lines essentially reemit the integrated stellar luminosity of galaxies, and thus are a direct and sensitive probe
of the young massive stellar population. Nebular lines include the widely used Hα line,
but also include Hβ, Pα, Brα, and Brγ for example. The primary advantage of nebular
emission as a star formation tracer is the high sensitivity and the direct coupling to the
massive SFR. Nearby galaxies can be mapped in high resolution at these wavelengths, and
Hα can be detected out to large redshift. Again, limitations include this method’s sensitivity
to extinction and dust, and the assumed IMF in the conversion between flux and SFR. The
effects of extinction can be minimized by comparing Hα measurements with those of the IR
recombination lines. The equation below relates the lineflux in the Hα line, L with the star
formation rate with a coefficient that is based on an assumed IMF, dust model and stellar
population library.
SF R(Msun yr−1 ) = 7.9 × 10−41 L(Hα)(ergss−1 )
(1.4)
Finally, the far-infrared (FIR) continuum is a useful metric for the measurment of star
formation within a galaxy. A significant fraction of a galaxy’s bolometric luminosity is
absorbed by dust and re-emited at redder wavelengths. Since the absorption cross section
of dust is peaked in the FUV, the thermal infrared can be a useful, albeit, indirect tracer
of star formation. Essentially, the contribution of young stars to the heating of the dust
in addition to the optical depth of the dust in the star formation contribute to the FIR
measurment. In the simplest case, the FIR luminosity measures the bolometric luminosity
of a given starburst- which is true in dense regions like nuclear starbursts. However, in
more general cases, the situation is not quite as simple: there is a contribution to the
FIR from the star formation, but also other, less dense contributions, from, for example,
the interstellar radiation field. For reasons such as the above, this indirect method should
be used in conjunction with others, unless the case being considered is an ideal, highly
12
starbursting case.
SF R(Msun yr−1 = 4.4 × 10−44 LF IR (ergss−1 )
(1.5)
As discussed above, some indicators work better than others, and some are limited to
only certain galaxy populations, or locations. For instance, using the UV continuum assumes
that SF has remained constant compared to the short lifetimes of the UV population, and
is very sensitive to extinction, up to 3 magnitudes (Kennicut 1998). Alternatively, Hα is
highly sensitive to nebular emission from massive stars but is sensitive to IMF choice and
assumes that the SFR is traced by ionized gas. Each of these indicators is likely influenced
by the global environment, such as merger rate and AGN activity. It is thus important to
explore the relationship between an SFR indicator and the true underlying SFR in order
to appropriately calibrate these indicators using simulations. Specifically, dwarf galaxies
are especially interesting environments in which to study star formation indicators: these
low-mass systems are characterized by burst of star formation rather than continuous or
active modes of star formation. Qualatatively, galaxies with bursty SFHs should imprint a
clear signiture on the distribution of SFRs.
1.1.4
Feedback
Feedback is the process by which energy is injected back into the ISM. Without this process,
star formation would be a perfectly efficient process: gas would easily cool, collapse and
form stars. However, large amounts of energy from stars is, in fact, transferred back to
the ISM as thermal and kinetic energy in the form of feedback, a key ingredient to any
successful galaxy simulation. The injection of energy into the ISM is what regulates star
formation.
When a star forms, it emits large amounts of high energy photons which starts the
destruction of it’s birth molecular cloud. The star evolves and travels along the main
sequence, and dies a violent death, which emits lots more ionizing photons. These photons,
heat, dissociate and ionize the surrounding medium in the form of supernovae and stellar
13
winds. These feedback events also enrich the ISM by depositing metals formed in these
stars, back into the ISM. The initial dump of energy from a supernova explosion results in
a shock in the surrounding material, which sweeps up the nearby gas. This wave of gas may
spark future star formation and also contributes to the turbulence of the ISM. Eventually,
the outward expansion of the supernova slows to the speed of the surrouding turbulence,
and the cloud dissipates. In short, the process of feedback regulates star formation, begining
with the destruction of a star’s birth cloud through to the death of the star in a violent
explosion injecting energy and turbulence into the ISM. Star formation and feedback are
inherently tied together not only with each other but also with the IMF which regulates
how many and what types of stars form, in addition to the supernova rate.
1.1.5
Simulating the Big Picture: Putting it All Together in Simulations
The interplay between star formation, the ISM and feedback and its affects on the large
scale properties of a galaxy is an incredibly difficult, detailed, and complex problem. Nbody + hydrodynamic simulations are one method of studying these complex systems. By
connecting gas cooling star formation and feedback we hope to trace the evolution of a
galaxy, including the following processes:
• Dark Matter and gas accrete into halos, along filaments or through mergers.
• Gas cools in these halos, forms molecular clouds and star formation begins.
• The galaxy increases in mass; more stars form, evolve and explode in feedback events
injecting metals and energy into the ISM. This suppresses star formation efficiency.
• Depending on the size of the halo (i.e., the depth of the potential well), supernovae
and other feedback sources can expell some gas from the galaxy: most, but not all of
this gas generally remains within the halo and returns to the disk after it has cooled,
redistributing star formation and slowing feedback.
14
• While locally feedback controls star formation, globally, star formation is controlled
by the accretion of gas into the galaxy and gas lost to the galaxy in stars. Up until
z = 2, gas can be accreted rapidly into halos and builds up. Afterwards, gas mass
decreases and SF decreases as well.
It is clear in this simple picture of galaxy evolution that it is imperative that simulations
include prescriptions for star formation, feedback, gas accretion and gas loss. This mandates
the use of baryonic physics in simulations. I have discussed techniques of following the
gas in simulations, but simulations generally also include models for physical processes
including star formation, supernovae feedback and metal enrichment. These recipes are
generally limited by the resolution of the simulation and some of these processes, including
star formation, happen on size scales smaller than the smallest resolution elements: either
the grid size or smoothing length, which then must be modeled semi-analytically for that
resolution element (subgrid modeling). As stated earlier, simulations have been successful in
simulation the formation and evolution of galaxies over a whole hubble time and reproduce
observed galaxy scaling relationships including but not limited to: the mass-metallicity
relationship, the Tully-Fisher relationship and as discussed in this thesis, the stellar to halo
mass relationship.
1.1.6
Limitations to Simulations and Possible Solutions
While simulations have had much success in describing the formation and evolution of
galaxies, they have their share of limitations. These limitations are mainly small scale
discrepancies between simulations and observations. Specifically, the traditional limitations
include: the missing satellite problem (i.e., simulations predict too many small halos that
are not observed), the cusp-core problem (high dark matter mass concentrations in centers
of halos), and the closely related angular momentum problem (a disproportionate amount
of low angular momentum material is driven to the center of halos resulting in smaller than
observed disks and overly massive bulges).
15
Figure 1.2 Plot of the mass-metallicity relationship in simulations, compared with observations. This plot shows the agreement between simulations (data points) and observations
(solid line). Plot from Brooks et al. (2007).
• Missing Satellite Problem: Most notable when comparing the number of Milky-Way
satellites to the number of satellites predicted by simulations of similar mass (Kaufmann, 1993; Klypin et al., 1999; Moore et al., 1999a). Even including observations of
ultra faint dwarfs (see e.g. Willman et al. (2004, 2005)), a high resolution simulation
will predict orders of magnitude more satellites than observed (Kuhlen et al., 2008;
Springel et al., 2008; Madau et al., 2008; Boylan-Kolchin et al., 2011b,a).
• Cusp-Core Problem: When comparing the dark matter mass profiles of simulated and
observed galaxies, simulations tend to have a cuspy profile while observations show
evidence of cored profiles (see e.g. Moore (1994); Burkert (1995); Oh et al. (2011a)).
16
Figure 1.3 Plot of cusp vs. core in collisionless simulations, compared with simulations that
include baryons. The black dot-dashed line shows the cuspy profile that represents the
overproduction of low angular momentum DM. The inclusion of baryons and the effect of
feedback shows that simulations can now reproduced cored profiles that are often observed.
Plot from Governato et al. (2012).
• Angular Momentum Problem: Simulations have a decreased disk mass, and increased
bulge mass (decreased high angular momentum material and increase low angular
momentum material). Historically, simulations could not retain a disk because of this
problem, especially during the course of mergers (Frenk et al. (1985); Bullock et al.
(2001)). With the ability now to retain disks, it is bulge size that becomes the worry:
17
specifically bulge to disk ratios (Guedes et al., 2011; Scannapieco et al., 2010; Brooks
et al., 2011).
These issues have been addressed in various ways over the years including those who
advocated for a different cosmology and those who advocate for different dark matter (warm
dark matter, self-interacting dark matter) over the traditional cold dark matter paradigm
(Bode et al., 2001). In a less drastic vein, simulations have addressed the above issues with
the addition of more extensive baryonic physics, which all focus on limiting the amount
of star formation in simulations- this solves the missing satellite problem and decrease the
baryonic mass within a galaxy bulge. These solutions include improved feedback methods
(stronger feedback, inclusion of AGN feedback and a UV background) (Quinn et al., 1996;
Gnedin, 2000; Benson et al., 2002; Somerville et al., 2008; Hopkins et al., 2011a). Increasing
feedback, specifically, has been shown as an effective way to reduce the amount of lowangular momentum gas at the center of galaxies, and can even more dramatically, cause
cored profiles at the centers of galaxies (through the interactions between baryons and dark
matter) (Governato et al., 2010; Brook et al., 2011; Pontzen & Governato, 2012b).
1.2
Outline
This thesis broadly encompasses the complex topic of star formation, specifically in Nbody+SPH cosmological simulations. Specifically this thesis tries to answer the following
questions:
• How do galaxies populate their dark matter halos? What affect does feedback and
baryonic physics have on the stellar to halo mass relationship?
• When and where are stars forming in galaxies? What is the structure of the ISM
where these stars are forming? How does this relate to the underlying IMF of these
systems?
18
• How well do observational star formation indicators trace star formation? What biases
do we see when comparing simulations to observations? What does this tell us about
the IMF in dwarf galaxies?
In Chapter 2, I describe my work comparing simulations to the observed stellar mass halo
mass (SMHM) relationship. I show that using an apples-to-apples approach in comparing
simulations to observations, we resolve the tension between simulations and observations.
I show effect of including baryons in a simulation on the dark matter halo masses and
the virial radius, as well as what happens when you apply observational mass estimates
to simulations. We find that dark matter halo mass decreases by 20% between a baryonic
and collisionless run, and that observations underestimate stellar mass by nearly 40%, when
compared to the true underlying stellar mass distribution.
In Chapter 3, I present an analysis of the star forming ISM in the bulge and disk of 4
simulated milky way mass galaxies. I show that the pressure of the star forming ISM in the
bulges of these galaxies is an order of magnitude higher than that of the star forming ISM
in the disks. We suspect that this pressure differential in the ISM is evidence for a varying,
time-dependent IMF which is more bottom heavy at early times, which implies a different
IMF forming the stars in the bulges of these galaxies, when compared to the disk.
In Chapter 4, I once again apply observational techniques to simulations of dwarf galaxies. Specifically, I examine the percieved tension between star formation indicators in the
UV and using recombination lines in Dwarf galaxies. I show that applying observational
measurements to simulated dwarfs yield consistent measurements of both these star formation indicators and explore the effect of resolution on star formation in these simulations.
Finally, in Chapter 5, I present the overall conclusions of the work presented in this
thesis and discuss future scientific studies based on this work, which I intend to pursue as
a post-doc.
19
Chapter 2
REPRODUCING THE STELLAR MASS/HALO MASS RELATION IN
SIMULATED ΛCDM GALAXIES: THEORY VS OBSERVATIONAL
ESTIMATES
Reproduced from Munshi et al. (2013b) with permission from the AAS.
We examine the present–day total stellar-to-halo mass (SHM) ratio as a function of halo
mass for a new sample of simulated field galaxies using fully cosmological, ΛCDM, high
resolution SPH + N-Body simulations. These simulations include an explicit treatment of
metal line cooling, dust and self-shielding, H2 based star formation and supernova driven
gas outflows. The 18 simulated halos have masses ranging from a few times 108 to nearly
1012 M⊙ . At z=0 our simulated galaxies have a baryon content and morphology typical of
field galaxies. Over a stellar mass range of 2.2 × 103 - 4.5 × 1010 M⊙ we find extremely
good agreement between the SHM ratio in simulations and the present–day predictions
from the statistical Abundance Matching Technique presented in Moster et al. (2012). This
improvement over past simulations is due to a number systematic factors, each decreasing
the SHM ratios: 1) gas outflows that reduce the overall SF efficiency but allow for the
formation of a cold gas component 2) estimating the stellar masses of simulated galaxies
using artificial observations and photometric techniques similar to those used in observations
and 3) accounting for a systematic, up to 30% overestimate in total halo masses in DM-only
simulations, due to the neglect of baryon loss over cosmic times. Our analysis suggests that
stellar mass estimates based on photometric magnitudes can underestimate the contribution
of old stellar populations to the total stellar mass, leading to stellar mass errors of up to
50% for individual galaxies. These results highlight that implementing a realistic high
density threshold for SF considerably reduces the overall SF efficiency due to more effective
feedback. However, we show that in order to reduce the perceived tension between the star
20
formation efficiency in galaxy formation models and in real galaxies, it is very important to
use proper techniques to compare simulations with observations.
2.1
Introduction
In the standard Λ Cold Dark Matter (ΛCDM) paradigm (White & Rees, 1978; Fall &
Efstathiou, 1980; Blumenthal et al., 1984; Dekel & Silk, 1986; White & Frenk, 1991), many
galaxy properties are expected to correlate with the mass of the galaxy’s host halo. In
particular, the stellar-to-halo mass relation (SHM), defined as the ratio of the stellar mass
(Mstar ) within a halo of total mass Mhalo within a given over-density (< ρ > /ρcrit =
200 in this work) is a robust estimator of the efficiency of gas cooling and star formation
(SF) processes over a wide range of halo masses (Somerville & Primack, 1999; Bower et al.,
2010). Both observational (Heavens et al., 2004; Zheng et al., 2007) and theoretical work
(Bower et al., 2012) suggest that the SF efficiency peaks at the scale of L⋆ galaxies and
declines at smaller and larger masses. On the low mass end, this decline is likely because
SF is suppressed by gas heating from the UV cosmic field (Gnedin, 2000; Okamoto et al.,
2008; Nickerson et al., 2011) and supernova (SN) heating with gas removal. At larger
masses, energy feedback from super-massive black holes (SMBHs) is thought to be the
dominant process responsible for lowering the SF efficiency (Bower et al., 2006; Croton,
2009; McCarthy et al., 2011; Johansson et al., 2012).
Recently, Moster et al. (2012, hereafter M12) and other groups (Vale & Ostriker, 2004;
Conroy et al., 2006; Mandelbaum et al., 2006; More et al., 2009; Guo et al., 2010; TrujilloGomez et al., 2011; Behroozi et al., 2013) used the Abundance Matching Technique (AMT)
and its variations (Yang et al., 2012) to derive a SHM relation of real galaxies. In its
simplest form AMT assumes a monotonic relation between the stellar mass function of
(real) galaxies and the underlying halo mass function. This relation is constrained by
matching the observed galaxy stellar mass function to the ΛCDM halo mass function from
N–body simulations. Similar works (Guo et al., 2010; Leauthaud et al., 2011, 2012) have
included constraints from lensing and used slightly different underlying cosmologies. This
21
approach has also been used to constrain the scatter in the SHM (Reddick et al., 2012)
by comparing the predicted spatial clustering of DM halos (Sheth et al., 2001; Reed et al.,
2007; van Daalen et al., 2012) with the observed abundances and clustering properties of
galaxy populations (Blain et al., 2004; Conroy et al., 2006; Reid et al., 2010). Additionally,
Behroozi et al. (2013) discuss the implications of the upturn in the faint-end slope of the
stellar mass function on the SHM relationship.
Several works have highlighted how uncertainties in the derived SHM relation depend
on a number of factors, some of them poorly known. For example the stellar masses of real
galaxies are inferred from optical and near–IR photometric measurements and/or resolved
spectra (Bell & de Jong, 2001; Kauffmann et al., 2003). This approach carries substantial
uncertainties and possible degeneracies (Bell & de Jong, 2001; Bell et al., 2003; Pforr et al.,
2012; Huang et al., 2012; Behroozi et al., 2010) as the observed spectral energy distribution of a galaxy is a function of many physical processes (e.g., stellar evolution, SFH and
metal-enrichment history, and wavelength-dependent dust attenuation, Panter et al., 2004).
Furthermore, as surveys often measure individual galaxy magnitudes within an aperture
based on a surface brightness cutoff, the mass of the stellar component could be systematically underestimated by at least 20% (Graham et al., 2005; Shimasaku et al., 2001) if
part of it is old (hence faint) and/or low surface brightness. Finally, the number density of
galaxies will be affected by incompleteness at the faint end of the galaxy luminosity function
(Dalcanton, 1998; Sawala et al., 2011; Geller et al., 2012; Santini et al., 2012). Separate
from worries over the stellar mass determinations are worries about the halo mass function.
While the halo mass function obtained in DM-only simulations is robustly constrained (Reed
et al., 2003; Springel et al., 2005b), it has recently (Sawala et al., 2012) been shown that
halo masses in DM-only simulations exceed those obtained in simulations including baryon
physics by up to 30%, introducing another systematic bias in the AMT, as a galaxy of a
given stellar mass is matched with a too massive halo, pushing the stellar-halo mass ratio
down. Taken together, the above caveats suggest that a better understanding of the connection between galaxy masses and the underlying halo masses could in principle be gained
22
by using realistic simulations of galaxy formation that directly include baryon physics, SF
and SN feedback.
While substantial progress has been made in creating galaxies from cosmological initial
conditions (Scannapieco et al., 2010; McCarthy et al., 2012; Sales et al., 2012; Stinson et al.,
2012; Johansson et al., 2012), several recent studies have pointed out a large discrepancy
between the SHM relation estimated for real galaxies and the one obtained in several numerical simulations of galaxy formation (Sawala et al., 2011; Guo et al., 2010). Simulations
have repeatedly shown that a lack of realistic SN feedback leads to overestimating star
formation as part of the general overcooling problem (Abadi et al., 2003; Governato et al.,
2007; Piontek & Steinmetz, 2011; Keres et al., 2011). Most simulations that overproduce
stars form galaxies that have large spheroidal components (Eke et al., 2001). Incremental
improvements based on more realistic SN feedback (Thacker & Couchman, 2000; Stinson
et al., 2006) led to simulations that formed galaxies with extended disks (Governato et al.,
2009; Brooks et al., 2011), but still substantially overproduced stars. Only recently, a new
generation of high resolution simulations demonstrated the impact of feedback at lowering
SF efficiency such that SF occurs only at high gas densities (Ceverino & Klypin, 2009; Governato et al., 2010; Guedes et al., 2011; Governato et al., 2012; Zolotov et al., 2012; Brook
et al., 2012). As SF is more efficient in dense gas clouds, feedback from these high density
regions generate outflows that simultaneously improve on several long standing problems
namely the substructure overabundance problem (Moore et al., 1998; Klypin et al., 1999;
Benson, 2010), reducing the B/D ratio in small galaxies by removal of low angular momentum baryons (Binney et al., 2001; Governato et al., 2010; Brook et al., 2011) and forming
DM cores by transferring energy from baryons to the DM (Mashchenko et al., 2006; Pasetto
et al., 2010; de Souza et al., 2011; Cloet-Osselaer et al., 2012; Macciò et al., 2012; Ogiya
& Mori, 2012; Pontzen & Governato, 2012a; Teyssier et al., 2012; Governato et al., 2012).
Forming stars in dense gas regions is a crucial step, as observations strongly support that the
spatially resolved SF is linked to the local H2 fraction (Bigiel et al., 2008; Krumholz et al.,
2009; Genzel et al., 2012), which only becomes significant at the density of star forming
23
regions, ≥ 10-100 amu/cm3 .
The relationship between gas density and H2 abundance can now be naturally implemented in simulations, and the creation and destruction of H2 can be followed consistently
(Gnedin et al., 2009; Christensen et al., 2012c), allowing much more realistic simulations
to be run, and establishes a physically motivated connection between SF and high density
(shielded) gas. This is indeed one of the major steps forward in the work presented here.
While feedback from SNe remains still poorly understood, its effects are being observed over
a large range of redshifts and galaxy masses (Martin, 1999; Wang et al., 2010). It is therefore important to evaluate a new set of high resolution simulations to test if outflows can
form galaxies with realistic observational properties that also reside on the SHM relation.
Relatively less attention has been given to comparing results from simulations with
observational estimates of the SHM relation in a consistent way. While some recent works
reported (Sawala et al., 2011; Avila-Reese et al., 2011; Piontek & Steinmetz, 2011; Leitner,
2012) an excess of stars formed in simulations, they compared the galaxy stellar and total
halo masses directly measured from simulations while those quantities are actually inferred
from the light distribution of real galaxies. A consistent approach has been already tried
with promising results in Oh et al. (2011b), where two simulated dwarfs were compared
with a set of galaxies from the THINGS survey, finding that the simulations have the same
central baryon and DM distribution as in the observational sample. In that work, stellar
and halo masses were obtained using photometric and kinematic data for both the simulated
and the real sample, finding excellent agreement (see Oh et al. (2011b) figure 5).
In this paper, we present a consistent comparison between the SHM estimated in Moster
et al. (2012) and a set of high resolution simulations spanning 5 orders of magnitude in stellar
masses. The evolution of galaxies is simulated at high resolution using Smoothed Particle
Hydrodynamics (SPH) in a cosmological context. We focus on how galaxies populate dark
matter halos ranging from 108 − 1012 M⊙ in a field environment and comparing results
with the estimates of the SHM from M12. The four most massive, MW–like galaxies have
similar resolution to the “Eris” galaxy (Guedes et al., 2011), while the smaller galaxies have
24
even better force resolution ( down to 65 pc and star particles as small as 450 M⊙). All
simulations include the effects of metal line cooling and H2 dependent star formation.
Our simulations include cooling, star formation, a cosmic UV background and form
galaxies with structural properties comparable to the real ones. The dataset utilized is
described in detail in Section 2 and has been analyzed in other papers showing that the
galaxies follow the Kennicutt-Schmidt law (Christensen et al., 2012c), have cored DM profiles similar to the observed ones (Oh et al., 2011b; Governato et al., 2012) and realistic
satellite populations (Zolotov et al., 2012; Brooks & Zolotov, 2012). Without any further
fine tuning, in this work we compare the same simulations to the SHM relation obtained
in M12, using an analysis technique comparable to the one used in the original paper to
estimate their stellar and halo masses. We show that simulations form realistic galactic
systems that also match the z=0 SHM of real galaxies over five orders of magnitude in
stellar mass.
The paper is organized as follows: in §2 we describe the details of our N-body simulations.
In §3 we compare results with the SHM predicted in M12. The results are discussed in §4.
2.2
The Simulations
The simulations used in this work were run with the N-Body + SPH code GASOLINE
(Wadsley et al., 2004; Stinson et al., 2006) in a fully cosmological ΛCDM context: Ω0 = 0.26,
Λ=0.74, h = 0.73, σ8 =0.77, n=0.96. The galaxy sample was selected from two uniform
DM-only simulations of 25 and 50 Mpc per side. From these volumes a few field–like
regions were selected and then resimulated at higher resolution using the ‘zoomed-in’ volume
renormalization technique (Katz & White, 1993; Pontzen et al., 2008). This technique allows
for significantly higher resolution while faithfully capturing the effect of large scale torques
that deliver angular momentum to galaxy halos (Barnes & Efstathiou, 1987). With this
approach, the total high resolution sample contains eighteen field galaxies, each halo resolved
by 5×104 to a few 106 DM particles within Rvir , defined as the radius at which the average
halo density = 200 × ρcrit .
25
The force spline softening ranges between 64 and 170 pc in the high resolution regions of
each volume and it is kept fixed in physical coordinates at z < 10. Star particles are formed
with a mass of 400-8000 M⊙ . The halo mass range covered by the simulations spans nearly
four orders of magnitude, from a few times 108 to 8 × 1011 M⊙ (peak velocities Vpeak =
10 to 200 km/sec), and stellar masses Mstar from 104 to a few 1010 M⊙ . As other works
have highlighted the importance of having a representative sample before drawing general
conclusions (Brooks et al., 2011; Sales et al., 2012; McCarthy et al., 2012), the halos in our
sample span a representative range of halo spin values and accretion histories (Geha et al.,
2006). Galaxies and their parent halos were first identified using AHF1 (Gill et al., 2004;
Knollmann & Knebe, 2009). The total halo mass (including DM, gas and stars) is defined
at a radius Rvir , defined as the radius at which the average halo density = 200 × ρcrit ,
consistent with M12. No sub-halos have been included in our sample, although the most
massive galaxies have a realistic population of satellites (Zolotov et al., 2012). More details
on this dataset and the properties of the satellite population are given in Governato et al.
(2012) and Brooks & Zolotov (2012).
2.2.1
H2 fraction, Star Formation and SN Feedback
In a significant improvement, this new set of simulations include metal line cooling (Shen
et al., 2010) and a dust dependent description of H2 creation and destruction by LymanWerner radiation and shielding of HI and H2 (Gnedin et al., 2009; Christensen et al., 2012c,
hereafter CH12). As in CH12, the star formation rate (SFR) in our simulations is set by the
local gas density and the H2 fraction; SF ∝ (fH2 ×ρgas )1.5 . A SF efficiency parameter, c∗ =
0.1, gives the correct normalization of the Kennicutt-Schmidt relation (the SF efficiency for
each star forming region is much lower than the implied 10%, as only a few star particles
are formed before gas is disrupted by SN winds). With the inclusion of the H2 fraction term
(see also Kuhlen et al., 2011), the efficiency of SF drops to zero in warm gas with T > 3,000
K. The simulations include a scheme for turbulent mixing that redistributes heavy elements
1
AMIGA’s Halo Finder, available for download at http://popia.ft.uam.es/AHF/Download.html
26
among gas particles (Shen et al., 2010). With this approach, the local SF efficiency is linked
to the local H2 abundance, as regulated by the gas metallicity and the radiation field from
young stars, without having to resort to simplified approaches based on a fixed local gas
density threshold (Governato et al., 2010; Kuhlen et al., 2011). Christensen et al. (2012a)
highlights the differences and improvements of H2-based star formation by comparing the
total amount, distribution and history of the gas and stars formed in simulations with and
without molecular hydrogen. The simulations assumed a Kroupa IMF and relative yields,
but observable quantities have been converted to a Chabrier IMF, for a direct comparison
with Moster et al. (2012).
As in previous works using the “blastwave” SN feedback approach (Stinson et al., 2006;
Governato et al., 2012), mass, thermal energy, and metals are deposited into nearby gas when
massive stars evolve into SNe. The amount of energy deposited amongst those neighbors
is 1051 ergs per SN event. Gas cooling is then turned off until the end of the momentumconserving phase of the SN blastwave which is set by the local gas density and temperature
and by the total amount of energy injected, typically ten million years. Equilibrium rates
are computed from the photoionization code Cloudy (Ferland et al., 1998a), following Shen
et al. (2010). A spatially uniform, time evolving, cosmic UV background turns on at z = 9
and modifies the ionization and excitation state of the gas, following an updated model of
Haardt & Madau (1996).
This feedback model differs compared to other “sub-grid” schemes (e.g., Springel &
Hernquist, 2003; Scannapieco et al., 2012) in that it keeps gas hydrodynamically coupled
while in galactic outflows. The efficient deposition of SN energy into the ISM, and the
modeling of recurring SN by the Sedov solution, should be interpreted as a scheme to
model the effect of energy deposited in the local ISM by all processes related to young
stars, including UV radiation from massive stars (Hopkins et al., 2011b; Wise et al., 2012a).
The SFHs of the galaxies in our simulated sample are bursty, especially those residing in
halos smaller than 1010 M⊙ . As discussed in Brook et al. (2011) and Pontzen & Governato
(2012a), a bursty SFH is necessary to remove low angular momentum baryons and create
27
the fast outflows able to transfer energy from baryons to the DM component and create
DM cores (Governato et al., 2010). These processes lead to realistic dwarf galaxies with
slowly rising rotation curves and typical central surface brightnesses 21 < µB,o < 23.5 (Oh
et al., 2011b). Outflows and the cosmic UV field progressively suppress star formation in
halos with total mass smaller than a few 1010 . Test runs verify that the effect of energy
feedback on suppressing SF is much larger than that of having a low H2 fraction-SF efficiency
(Christensen et al., 2012c).
2.2.2
Star formation efficiency as a function of galaxy halo mass
As a result of SF feedback processes, the SF efficiency is greatly reduced over the whole mass
range of our simulated sample. The smallest galaxies in our sample turn only ∼0.01% of
their primordial baryons into stars. The more massive galaxies in our sample turn ∼30% of
their primordial baryon content into stars, but we demonstrate in the next section that using
stellar masses based on photometry reduces this efficiency to an apparent 10%. Feedback
expels about 70% of the gas to outside of Rvir in dwarf galaxies with vc ∼ 40-50 km s−1 .
Larger galaxies retain a larger fraction of their original baryons, while the smallest galaxies
lose an even larger fraction of baryons due to gas heating by the cosmic UV background,
which further reduces their SF (see Figure 1).
To evaluate the effect of the adopted SF model on the resulting SF efficiency of our galaxy
sample, galaxies were re-simulated using the SF approach used in our older works. These
reference runs adopt a lower density threshold for SF, 0.1 amu/cm3 . As discussed in several
works (Governato et al., 2010; Guedes et al., 2011), a low density threshold makes SF more
diffuse and locally less efficient. In this scenario, typical of low resolution simulations where
star forming regions cannot be resolved, the amount of SN energy per unit mass delivered to
gas particles is effectively lowered, making feedback much less effective at suppressing SF.
While the galaxies in the low threshold sample have realistic disk sizes (Brooks et al., 2011)
and the total amount of energy released into the gas is actually a few times larger, they
contain many times more stars (see Figure 2) and overproduce stars compared to the SHM
28
relation. The comparison between the amount of stars formed in the old and the new runs
(Figure 2) demonstrates that the large decrease in the SF efficiency in the new simulations
(as much as a factor of 15) is due to the improved implementation of SF in dense/H2 rich
regions (see also Governato et al., 2010). This lower SF efficiency goes a long way toward
reconciling simulations of galaxy formation with current estimates of the SHM relation.
2.2.3
The Baryon Content of the Simulated Galaxies
The more massive galaxies in our sample are disk dominated, transitioning to irregular
galaxies below ∼1010 M⊙ . The outflows (and the UV cosmic field in halos below ∼109 M⊙ )
significantly lower the baryon fraction of the host halos, with a strong trend of lower baryon
fractions at smaller halo (stellar) mass (see Figure 1). However, because the fraction of
remaining gas turned into stars is low at all galaxy masses and especially in dwarfs, the
galaxies in our sample have relatively high cold gas to stellar mass ratios, typical of real
galaxies over the tested range. In Figure 3 we compare the cold gas (HI) content of the
simulated sample to the nearby HI surveys ALFALFA (Giovanelli et al., 2005) and SHIELD
(Cannon et al., 2011). The observed dataset includes only galaxies closer than the Virgo
cluster, for a better comparison with our sample of field galaxies. With the caveat that
selection effects can still play a role, there is very good agreement between simulations and
observations.
The HI masses are directly measured from the simulations, where HI and H2 2 abundances
are calculated on the fly. Magnitudes are measured in the SDSS r-band.
We verified that the cold gas fraction in the comparison runs adopting a low density
threshold for SF is about a factor of ten lower at all halo masses. Lower resolution simulations of small mass systems have often reported the formation of relatively gas poor galaxies
(Governato et al., 2007; Colı́n et al., 2010; Avila-Reese et al., 2011; Sawala et al., 2011).
Low gas content in simulated low-z galaxies is likely due to a high efficiency of gas to stars
2
note that H2 masses are small and, while neglected in Figure 3, contribute little to the overall cold gas
mass
29
conversion (Piontek & Steinmetz, 2011) and/or to an excessive loading factor of the SN
winds.
2.3
The Stellar Mass to Halo Mass Relationship
Once individual galaxies in our sample have been identified with AHF, the Stellar Mass
- Halo Mass ratio for each galaxy can then be obtained. The definition of “halo mass”
includes all DM and baryons within an overdensity of 200, but not the mass associated with
individual satellites (a few % of total at the most). All stars not in satellites, but within
R200 are associated with the central galaxy in the halo. This simple approach is similar to
what has been done in several previous works (Sawala et al., 2011; Brook et al., 2012) and
similar to what has been used in previous comparisons between simulations and the SHM
relation obtained using the AMT (Guo et al., 2010; Moster et al., 2012). Our new sample of
simulated field galaxies (open circles in Figure 4) follows closely the shape and normalization
of the present day SHM relation presented in M12 over the 109 –1012 M⊙ halo mass range
2 . This is a large improvement over most published works
(solid line), with Mstar ∝ Mhalo
and confirms results on smaller samples (Governato et al., 2010; Oh et al., 2011b; Guedes
et al., 2011; Brook et al., 2012) that adopted similar SF and feedback recipes.
Clearly, an approach where a) SF is limited to dense, H2 -rich gas clouds (a highly
correlated situation, see Christensen et al., 2012c) and b) feedback is hydrodynamically
coupled to outflows significantly reduces the SF efficiency and the present day stellar mass
in galaxy sized halos over a wide mass range. Our simulations show that both a) and b)
are alone not sufficient, but the combination is sufficient. We have first verified that in the
absence of feedback the SF efficiency remains high even if the consumption rate of gas is
long, as over the course of a Hubble time most cold gas within the galaxy eventually turns
into stars. Moreover, lack of SN feedback fails to remove the low angular momentum gas,
originating galaxies with an excessive spheroidal component (Governato et al., 2010; Brook
et al., 2011). Similarly, if the identical SN feedback recipe used in this work is applied to
simulations where SF is allowed in cold, but relatively low density gas (e.g 0.1 amu, as often
30
adopted in the past), it fails to significantly lower the overall SF efficiency. In our sample,
the mass in stars formed by z = 0 in galaxies with H2 /high density regulated SF is lower
by as much as fifteen compared to simulations of the same halos adopting a lower density
threshold. The overproduction of stars in the low threshold runs occurs even when metal
lines cooling is neglected.
In summary, while the SF efficiency in the high threshold simulations is lower, the cold
gas content is similar to that observed in real galaxies (see §2.3). Hence, a low SF efficiency
was not obtained by simply increasing the feedback strength and “blowing away” all the
baryons. Combined, these results show that adopting a more realistic description of where
stars form and how feedback regulates SF leads to realistic simulations of galaxies.
However, for a meaningful comparison with observations and the SHM inferred in M12,
it is important to infer both the stellar and halo mass from the simulations using the same
techniques as the observations. This additional step is necessary as simulations directly
measure the mass distribution, while observations infer the stellar mass from the light distribution. Below we show that this more accurate approach affects the results substantially.
We will provide, for the first time, an accurate comparison with the present day SHM
relation obtained from observational data. We used the following procedure:
• Magnitudes based on the age and metallicity of each star particle were derived using
the Starburst99 stellar population synthesis models of Leitherer et al. (1999) and
Vázquez & Leitherer (2005), adopting a Kroupa (2001) IMF.
• For each simulated galaxy, Petrosian aperture magnitudes (Blanton et al., 2001; Yasuda et al., 2001) were obtained in the r band. This step is necessary as observations
are limited by the surface brightness of the target galaxy dropping below the sky
brightness. This systematic bias underestimates the amount of light associated with
individual galaxies, and it is estimated to be of the order of 20% for real galaxies (Dalcanton, 1998; Blanton et al., 2001). As our galaxies have light profiles that closely
mimic those of real galaxies (Brooks et al., 2011), applying this constraint is appro-
31
priate. We verified that the amount of light lost is similar to that estimated for
observational samples.
• The stellar mass of each galaxy was then estimated based on its B-V color and V total
magnitude, assuming a Salpeter IMF to be consistent with adopting the same fitting
formula as in Bell & de Jong (2001), namely LV = 10−(V −4.8)/2.5 and then Mstar =
LV × 10−0.734+1.404×(B−V ) . We then utilize a conversion from Salpeter to Chabrier
IMF to remain consistent with M12. We find that this procedure systematically
underestimates the true stellar masses (by summing all star particles within Rvir not
in satellites) of galaxies by 20-30%. This result extends over the whole range of galaxy
masses. We find that the specific criteria adopted in Bell & de Jong (2001) tends to
underestimate the contribution from old (i.e., high M/L) stellar populations.
• The halo mass for each simulated galaxy was measured re-running each simulation as
DM–only, matching halos between the two runs and counting all mass with Rvir (again
defined as the radius within which the average overdensity is < ρ >= 200ρcrit ). This
step is necessary to follow the procedure adopted in Moster et al. (2012), where halo
masses were obtained from a large cosmological simulation that did not include gas
physics. This procedure avoids a subtle, but significant and systematic bias between
the total halo mass measured in DM-only runs vs those of the same halos in simulations
that include gas physics and feedback. In the latter simulations, feedback can remove
a significant fraction of the baryons from the halo, decreasing the total mass within a
fixed physical radius. The virial radius, if defined at a fixed overdensity, then shrinks,
leading to a smaller total halo mass. The decrease in Mvir varies with mass, as it
depends on the amount of baryons lost, but it can be significant, up to 30%. Since
the lowest mass galaxies in our sample have lost the most baryons (in winds and UV
background heating), they can experience a decrease of ∼30% in halo mass. At the
high mass end, where galaxies retain most of their baryons, the simulated galaxies
still see a change of 5-10% in total halo mass compared to the DM-only run (where
32
obviously no baryon mass loss is possible). These results are consistent with estimates
in (Sawala et al., 2012). Sawala et al. (2012) also interprets this shift as a systematic
reduction in the matter infall rate. As even small amounts of baryons are removed, the
gravitational attraction on surrounding material decreases, leading to a decrease in the
infall rate of both gas and DM and, overtime, to a smaller halo mass. Neglecting this
effect results in moving the simulations datapoints to the left, away from the SHM
relation inferred using DM-only runs. This bias is particularly noticeable at small
galaxy masses, where the SHM relation is steeper.
In Figure 4, the black solid dots show results from our simulations dataset, but using the
procedure outlined above, which closely matches that adopted in M12. The normalization
of the SHM is ∼ 40% lower than that inferred using the simulations quantities at face value
and closer overall to the SHM relation of M12. There is very good agreement between the
SHM inferred from the ‘artificial observations’ and the fit in M12 over at least 4 orders of
magnitude in halo mass. The mass of the brightest galaxies (close in mass and morphology
to Milky Way analogues) goes from being higher to being slightly lower than the SHM.
This result confirms that the “blastwave” feedback implementation is able to reduce the SF
efficiency not only at small halo masses but also in present day halos around 1012 M⊙ (see
also Guedes et al., 2011).
In Figure 5a we show the total halo mass ratio between the simulations that include
baryons and SF vs the DM-only ones. As discussed above, halos in DM-only runs are
consistently (and significantly, about 30%) more massive.In Figure 5b we show the ratio
between the stellar masses obtained using a combination of Petrosian Magnitudes with Bell
& de Jong (2001) M/L ratios (closely following M12) and stellar masses derived directly
from the simulations. A systematic bias of about 50% is evident across the whole mass
range. The results from this section highlight the importance of a careful comparison
between simulations and observations. This approach further reinforces our findings that
the modeling of SN feedback greatly reduces the tension between the present time SHM
relation inferred from the Abundance Matching Technique for halos with total mass <
33
1012 M⊙ and the predictions from hydrodynamical simulations in a cosmological context.
Without fine tuning, this better agreement also preserves the morphology of the galaxies
formed, with disk-dominated massive galaxies and low mass irregulars that are gas rich.
In future work, we plan to extend this approach to higher redshifts, using the appropriate
criteria to measure stellar masses in high-z galaxies (e.g., Pforr et al., 2012; Maraston et al.,
2012).
2.4
Conclusions
We have measured the SHM (stellar mass – halo mass) relation for a set of field galaxies
simulated in a ΛCDM cosmology and compared it with the redshift zero predictions based
on data from the SDSS and the Abundance Matching Technique described in M12. The
comparison revealed very good agreement in normalization and shape over five orders of
magnitude in stellar mass. The new simulations include an explicit description of metal
lines cooling and H2 regulated SF, and SN driven outflows. The combination of SF driven
by the local efficiency of H2 and outflows reduce the overall SF efficiency over the whole
Hubble time by almost an order of magnitude compared to older simulations, with resulting
Mstar ∝ M2halo . While a large fraction of baryons is expelled, especially in halos smaller
than 1011 M⊙ , the resulting galaxies have an HI content comparable to those inferred by
local surveys, namely ALFALFA and SHIELD. The same galaxy set has a cored central DM
density distribution, similar to observations of real galaxies (Governato et al., 2012; Brooks
& Zolotov, 2012).
This agreement between simulations and observational data is due to two systematic factors: 1) An implementation of SF that relates the SF efficiency to the local H2 abundance
in resolved star forming regions, resulting in localized feedback that significantly lowers the
SF efficiency and 2) “observing” the simulations to properly compare them to observational
estimates of the SHM relation. This approach involved creating artificial photometric light
profiles of the simulated galaxies and estimating stellar masses based on aperture magnitudes. Importantly, it also requires coupling the stellar masses to halo masses derived
34
from DM-only simulations, rather than the baryonic simulations. Our analysis shows that
adopting photometric stellar masses contributes to a 20-30% systematic reduction in the
estimated stellar masses. Stellar mass estimates based on one band photometric magnitudes are likely to underestimate the contribution of old stellar populations (reflecting the
larger contribution to the total flux coming from younger stars). This systematic effect is
further exacerbated by the use of aperture based magnitudes, adding another 20-30% due
to neglecting the contribution of low surface brightness populations. Finally, a third systematic effect comes from a difference in halo masses in collisionless (DM-only) simulations
vs simulations including baryon physics and outflows. Baryon mass loss makes halo masses
smaller by up to 30% when calculated at the same overdensity (200 in our paper and M12).
The effect of removing these biases is to move the simulation points in Figure 4 further
lower and to the right, closer to the SHM.
Notwithstanding the improvements described in this and other recent works, further
adjustment to our numerical schemes to model SF and feedback processes are most likely
required, as more observational constraints become available and our understanding of SF
improves. In future work we plan to extend our analysis of the stellar mass – halo mass
relation to higher redshifts and larger galaxy masses. The results presented in M12 point
to a possible discrepancy between the shape of the star formation history of real galaxies vs
the simulated ones. Given the difficulty to obtain robust estimates from faint and distant
galaxy samples, we expect that the approach outlined in this work, i.e. creating artificial
observations to more directly compare simulations with observations, will play an important
role.
2.5
Supplemental Analysis
In this section I present extra analysis not included in the published paper regarding the
stellar mass to halo mass relation in our simulations. Specifically, I compare our abundance
matching results to the results from Behroozi (2010). Note that the Behroozi results show
an upturn in the SHM relationship toward lower masses: this is due to the fact that this
35
analysis includes corrections for incompleteness in the faint end of the luminosity function,
and adjusts the matching between the halo mass function and the luminosity function to
adjust for small galaxies missed by surveys. We also examine the affect of baryons on dark
matter halos by comparing virial radii and virial masses of runs completed with baryons
and those that are collisionless. In figures 2.8 and 2.9, we verify that for our galaxies with
halo masses greater than 1010 M⊙ that the change in halo mass can be explained entirely
by a change in virial radius that results from a new overdensity after baryons are lost. In
halos smaller than this, the smaller virial radius in the SPH run cannot completely account
for the change in halo mass, and it appears that the lowered mass of these halos also leads
to lower accretion. We also compared the DM mass between a baryonic and collisionless run
and we can see that the DM mass within the same radius is the same in each run for halos
more massive than 1010 solar masses. In these more massive galaxies, the change in the
halo mass can be entirely accounted for by the change in virial radius. In the lower masses,
there’s some indication that a small fraction of DM really isn’t within the same radius, and
that there might be a lowered accretion rate in these halos. In comparing the virial radii,
we compared total mass. For the SPH runs, we added back in the baryonic mass that had
been lost. The results are consistent with the DM mass plot, where the larger halos are
explained entirely due to the change in virial radius, and lower mass halos indicating some
small change in accretion within that radius.
In Figure 2.11, we also show that as halo mass increases, star formation efficiency increases by examining as a function of halo mass, the ratio of stellar mass to halo mass.
Note that low mass galaxies are very inefficient at forming stars, and that as halo mass
increases, star formation efficiency increases. Additionally, we also examine the efficacy of
using M/L ratios to estimate stellar mass in Figure 2.10. I created “images” of the galaxy
that contain different portions of the stellar mass of the galaxy: each image (at each formation time point) contains newer and newer stars, i.e., as the points increase along the
Tform axis, they contain and increasing amount of younger stellar population. For each
image that I generated, I measured the stellar mass, directly from the simulation (actual)
36
and stellar mass measured by using mass to light ratios (measured). The ratio of the two
quantities tells us what amount of the total stellar mass is recovered by using M/L ratios
(using B-V color) to estimate mass, for a given stellar population. As we include younger
stellar populations, we note that the mass recovered by stellar populations becomes better
which implies that using M/L ratios to estimate stellar mass is an underestimate as it does
not recover the old stellar population.
37
Figure 2.1 Baryonic fraction with respect to the cosmic ratio, for simulated field galaxies as a
function of stellar mass, measured at z=0. Circles are the “direct from simulation” results,
including all gas and stars within R200 . Triangles are the “observable” baryon fractions,
includind all stars and all the ‘observable’ cold gas (defined as 1.4 × (HI+H2), within R200 ).
The empty symbols are galaxies with no observable gas (cold gas mass < 100 M⊙ ). Symbol
sizes represent the different mass resolutions of galaxies in the sample. Smaller symbols are
the higher mass resolution by a factor of 8 when compared to the larger symbols. Galaxies
below 108 M⊙ lose a significant fraction of baryons due to heating from the cosmic UV
background and SN feedback.
38
Figure 2.2 The stellar mass ratio between the galaxies simulated with the old ‘low density SF
threshold’ and the new sample. In the new sample SF is regulated by the local abundance
of molecular hydrogen, resulting in feedback significantly lowering the total SF efficiency.
All quantities as measured directly from the simulations.
39
Figure 2.3 The cold gas mass as a function of stellar mass. Simulations vs. SHIELD
and ALFALFA data. The HI mass of each galaxy in the simulated sample is plotted vs
the SDSS r-band magnitude and compared to two samples from nearby surveys. Red
solid dots: simulations. Diamonds: ALFALFA survey. Asterisks: SHIELD survey. While
feedback removes a large fraction of the primordial baryons, the simulated galaxies have
a high gas/stellar mass ratio, comparable to the observed samples. Most of the cold gas
resides within a few disk scale lengths from the simulated galaxy centers. Figure courtesy
of A. Brooks.
40
Bell & DeJong Masses
Sim Masses
log Mstar (Msun)
10
8
6
4
8
9
10
11
log Mhalo (Msun)
12
Figure 2.4 The Stellar Mass vs Halo Mass. Black Solid Dots: The SHM relation from our
simulations set with stellar masses measured using Petrosian magnitudes and halo masses
from DM-only runs. This procedure mimics the one followed in M12. Open Dots: Unbiased
stellar masses measured directly from the simulations. Solid Line: Observational results
from M12. Symbol sizes represent the different mass resolutions of galaxies in the sample.
Smaller symbols are the better mass resolution by a factor of 8 when compared to the larger
symbols.
41
1.2
Mhalo,SPH/MDMO
1.1
1.0
0.9
0.8
0.7
0.6
0.5
8
9
10
11
log MDMO
12
10
11
log MDMO
12
1.4
Mstar,pet/Mstar,full
1.2
1.0
0.8
0.6
0.4
0.2
0.0
9
Figure 2.5 Top Panel: Halo mass ratio of galaxies in runs with baryons and SF vs DMonly runs. Individual halos in DM–only runs are typically 30% more massive than their
counterparts in simulations with gas physics and SF. The effect is smaller in more massive
halos, where baryon loss due to feedback is less (see also Sawala et al. (2012)). The dashed
horizontal line marks the ratio if halos had a 100% baryon loss. Middle Panel: Estimated
vs. True Stellar Mass as a function of halo mass. The stellar mass using artificial Petrosian
magnitudes and measured using the photometric method in (Bell & de Jong, 2001) vs the
“true” Stellar mass measured directly from the simulations. Stellar masses measured using
the photometric method in (Bell & de Jong, 2001) in combination with the flux loss from
applying the petrosian magnitudes are underestimated by about 50% across the range of
galaxy masses in our study.
42
Mstar,pet,color/Mstar,pet,sim
1.4
1.2
1.0
0.8
0.6
0.4
0.2
0.0
9
10
11
log MDMO
12
Figure 2.6 Stellar mass within Petrosian radius using colors vs straight from simulations.
This highlights the contribution of using observational techniques in the underestimation of
stellar mass, after the Petrosian radius is applied. The use of a fixed aperture underestimates
mass by a further 10-20%, as previously estimated in Blanton et al. (2001)
43
Bell & DeJong Masses
Sim Masses
log Mstar (Msun)
10
8
6
4
8
9
10
11
log Mhalo (Msun)
12
Figure 2.7 The Stellar Mass vs Halo Mass. Black Solid Dots: The SHM relation from our
simulations set with stellar masses measured using Petrosian magnitudes and halo masses
from DM-only runs. This procedure mimics the one followed in M12. Open Dots: Unbiased
stellar masses measured directly from the simulations. Solid Green Line: Observational
results from M12. Solid Red Line: Results from Behroozi 2012.
44
Figure 2.8 Comparison of dark matter halo mass between a baryonic run and a collisionless
run. This figure shows that in the larger halos, the change in virial mass can be accounted
for just by the change in overdensity after baryons are lost. However, in smaller halos, this
is not sufficient: these halos must also suffer from reduced accretion, thus the scatter below
1. Figure courtesy of Alyson Brooks.
45
Figure 2.9 Comparison of dark matter halo virial radius between a baryonic run and a
collisionless run. This figure, like the previous, shows that the change in virial radius is, in
larger halos, due to the change in overdensity due to baryonic physics. In the smaller halos,
this is not the case and these halos must have reduced accretion to account for the change
in virial radius. Figure courtesy of Alyson Brooks.
46
1.2
1.0
measured/actual
0.8
0.6
0.4
0.2
0.0
0.0
0.1
0.2
Tform
0.3
0.4
Figure 2.10 Anyalysis of mass recovered as a function of stellar population age when using
M/L ratios to estimate stellar mass. In this figure, we use B-V color, as in the analysis for
the SMHM relation. The first point is so high because the stellar population is so old that
the B-V colors are undefined. The x-axis is time in simulation units, ranging from 0 to 13.7
Gyrs.
.
47
Mstar/Mhalo
0.3
0.2
0.1
0.0
7
8
9
10
Mhalo (MΟ•)
11
12
Figure 2.11 Stellar mass in a given halo mass. This figure shows how the star formation
efficiency increases toward milky way halo masses.
.
48
Chapter 3
THE PRESSURE OF THE STAR FORMING ISM IN
COSMOLOGICAL SIMULATIONS
Reproduced from Munshi et al. (2013a) with permission from the AAS.
We examine the pressure of the star-forming interstellar medium (ISM) of Milky-Way
sized disk galaxies using fully cosmological SPH+N-body, high resolution simulations. These
simulations include explicit treatment of metal-line cooling in addition to dust and selfshielding, H2 based star formation. The 4 simulated halos have masses ranging from a few
times 1010 to nearly 1012 solar masses. Using a kinematic decomposition of these galaxies
into present-day bulge and disk components, we find that the typical pressure of the starforming ISM in the present-day bulge is higher than that in the present-day disk by an
order of magnitude. We also find that pressure of the star-forming ISM at high redshift is
on average, higher than ISM pressures at low redshift. This explains the why the bulge forms
at higher pressures: the disk assembles at lower redshift, when the ISM is lower pressure
and the bulge forms at high redshift, when the ISM is at higher pressure. If ISM pressure
and IMF variation are tied together as suggested in studies like Conroy & van Dokkum
(2012), these results could indicate a time-dependent IMF in Milky-Way like systems, as
well as a different IMF in the bulge and the disk.
3.1
Introduction
The origin of the stellar initial mass function (IMF) is paramount to our understanding of
star formation, stellar evolution and feedback and galaxy formation. The IMF influences
most of the observable properties of both stellar populations and galaxies. Detecting variations of the IMF will provide deep insights into the process by which stars form including
but not limited to: the origin of the stellar mass scale, the effects of metallicity and environ-
49
ment and the energetics of feedback. Additionally, the IMF is a key ingredient into a huge
range of models of all the above phenomena, and a necessary assumption when deriving
physical parameters from observations. Despite being such a vital ingredient, the origin
and variations of the IMF still remain poorly understood.
In particular, of critical importance, is the question of whether the IMF is universal or
whether the IMF is sensitive to the initial conditions of star formation- i.e., the structure of
the ISM in which the stars are forming (see e.g. Kroupa et al. 2011). Growing observational
evidence suggests that the high mass behavior of the IMF is uniform, including observations
of the IMF in the Magellanic Clouds (?Bastian et al., 2010; Chabrier, 2003). However at
the low mass end, there are many indications, both observationally and theoretically, that
there may be a variation in the IMF. For example, Conroy & van Dokkum (2012) and
van Dokkum & Conroy (2011) show that the IMF in these systems is bottom heavy using
gravity sensitive absorption lines in the cores of giant elliptical galaxies. This has also been
independently suggested by kinematic and lensing data (Treu et al., 2010; Cappellari et al.,
2012; Dutton et al., 2013). As these systems formed their stars at high redshift, these studies
give us insight into the time-evolution of the IMF. Observationally, Conroy & van Dokkum
(2012) show that the mass to light ratios of spheroidal systems indicate a more bottom
heavy IMF at higher pressures, and at higher SFRs. This indicates that ISM pressure and
the intensity of star formation are both key in understanding how and where stars formand whether or not the IMF is varying.
Despite the importance of the IMF, a universally agreed upon, fully cosmological, physical theory of its origin and variation with environment has yet to be found: rather, many
competing models exist. In particular, explaining the evolution of the IMF has been a challenge for theoretical models, especially in ’normal’ systems. In general, these studies predict
large IMF variations with local thermal Jeans Mass (Moore et al., 1999b; Larson, 2005),
Mach number of the star-forming ISM, or the distribution of densities in a supersonically
turbulent ISM (Padoan et al., 1997, 2012; Hennebelle & Chabrier, 2008; Hopkins, 2012b,a).
Most theoretical models offer explanations of IMF variation in more extreme conditions-
50
ULIRG, nuclear starbursts owing to the extreme mergers and large gas inflows (Kormendy
& Sanders, 1992; Hopkins et al., 2008; Hopkins, 2013; Narayanan & Davé, 2012; Narayanan
& Hopkins, 2013) but many predict a top-heavy scenario, contradictory to the observational
evidence. Furthermore, Krumholz (2011) shows that the critical mass, i.e. the fragmentation mass of a collapsing star-forming cloud, is dependent on the metallicity and pressure of
the cloud itself. In a toy model, Weidner et al. (2013) suggest a time dependent IMF with a
top heavy IMF slope followed by a prolonged bottom heavy slope which will bring the ISM
pressure, temperature and turbulence into states that will drastically alter fragmentation
of the gas, to explain observations.
In this letter, we perform SPH+N -body simulations of four medium-mass galaxies to
directly explore the star-forming ISM in a cosmological context. This study is unique in that
SPH simulations, being particle based, allow us to follow the full thermodynamic history
of the gas. Furthermore, the resolution and star formation recipe in these simulations
allow us to begin to probe the density structure of a more realistic star-forming ISM, in a
cosmological setting. Here we specifically focus on a comparison between the star forming
ISM of stars that make up the present-day bulge and those that make up the present-day
disk. We find that stars form at ISM pressures an order of magnitude higher in the bulge
than those in the disk, on average. Additionally, we find that at early times both bulge and
disk stars form in a high pressure ISM. Finally, we find that differences in the formation
radius of bulge and disk stars are not responsible for the different pressures. In short, we
show that there ISM pressure varies with time, which could imply that the IMF varies with
time as well.
3.2
The Simulations and Analysis
The simulations used in this work were run with the N-Body + SPH code GASOLINE
(Wadsley et al., 2004; Stinson et al., 2006) in a fully cosmological ΛCDM context: Ω0 =
0.26, Λ=0.74, h = 0.73, σ8 =0.77, n=0.96. Using the ‘zoomed-in’ volume renormalization
technique (Katz & White, 1993; Pontzen et al., 2008),we selected from uniform DM-only
51
simulations field–like regions which we then resimulated at higher resolution. This set of
simulations includes metal line cooling (Shen et al., 2010) and tracks non-equilibrium H2
abundances (Christensen et al., 2012d, hereafter CH12). As in CH12, the star formation
rate (SFR) in our simulations is set by the local gas density and the H2 fraction; SF ∝ (fH2
c∗ ×ρgas )1.5 , with c∗ = 0.1. Star formation is limited to gas with density greater than 0.1
amu/cc and temperature less than 3000 K, although the dependency of star formation on
the H2 abundance makes these limitations largely inconsequential. Tests of the convergence
of star formation histories for this star formation perscription are described in CH12.
The full details of our SN feedback “blastwave” approach are described in several papers
including, Stinson et al. (2006); Governato et al. (2012). As massive stars evolve into SN,
mass, thermal energy and metals are deposited into nearby gas particles, with energy of 1051
ergs per event. Gas cooling is shut-off until the end of the snow-plow phase as described
the Sedov-Taylor solution, typically ten million years. We also include gas heating from
a uniform, time evolving UV cosmic background, which turns on at z = 9 and modifies
the ionization and excitation state of the gas, following the model of Haardt & Madau
(1996).The efficient deposition of SN energy into the ISM, and the modeling of recurring
SN by the Sedov solution, should be interpreted as a proxy we have tuned to model the
effect of processes related to young stars and to model the effect of energy deposited in the
local ISM including UV radiation from massive stars (Hopkins et al., 2011b; Wise et al.,
2012b). The simulations also include a scheme for turbulent mixing that redistributes heavy
elements among gas particles (Shen et al., 2010). These feedback, star formation, and
ISM parameters in simulations of the same resolution produced galaxies with realisticallyconcentrated bulges (Christensen et al., 2012b). Furthermore, it is important to note that
because the simulations are tuned to produce realistic present-day galaxies (see e.g. Munshi
et al. (2013b)), with correct surface densities, the mean ISM pressure in these simulations
should be approximately correct, even if the feedback model does not include processes
specifically related to young stars, stellar winds and radiation pressure (Hopkins et al.,
2011b; Hopkins, 2013).
52
Galaxy Mhalo (M⊙ )
Gas
Par-
Name
ticle
Mass
Softening
(pc)
(M⊙ )
h986
1.9 × 1011
3900
115
h277
6.8 × 1011
3900
115
h258
7.7 × 1011
3900
115
h239
9.1 × 1011
3900
115
Table 3.1 Description of simulations utilized in this analysis.
We have simulated four different disk galaxies, at high resolution, described in Table
1. We dynamically decompose our disk galaxies based on cuts in angular momentum and
energy(Scannapieco et al., 2011; Governato et al., 2009). Each star particle at z = 0 is traced
back to the gas particle from which it formed in order in order to sample the properties of the
ISM from which each component formed. Using the cold gas in the central few kiloparsecs
of he galaxy, a star particle is established as disk when its specific angular momentum (jz ) is
a large fraction of the angular momentum of a circular orbit with the same binding energy,
i.e., jz /jc > 0.8. Using the potential of the entire matter distribution (dark, gas and stars),
we determine the total energy for each particle and subsequently it’s angular momentum.
For the bulge and halo stars, star particles are identified based on their radial orbits and
their binding energy: bulge stars have higher binding energies than halo stars. Furthermore
bulge and halo are also distinguished by the radius where the spheroid mass profile changes
to a shallower slope. We checked the stability of our kinematic decomposition across three
simulation outputs in time- ie over the course of 100 Myr which approximates the dynamical
time of the systems. We compared the results of our analysis in the case of the strictest
definition that particles must be classified as the same component over a whole dynamical
time to the weakest definition that particles need only be classified a component at z = 0
and found that the resulting trends remain unchanged.
Using full information from the simulations (kinematics, ages, metallicity), we have
53
traced the density, temperature, velocity dispersion and pressure of the ISM in which the
stars of each component form. Being particle based, the SPH approach of our simulations
allows us to follow in detail the thermodynamical history of the gas through cosmic times,
without resorting to additional ’tracer elements’ (Genel et al., 2013). This allows us to
examine not only where, but also in what environment stars are forming in our simulated
Milky-Way like galaxies. In our analysis, we define pressure in very simple terms: P = nkb T .
We get temperature and density by tracing each star particle belonging to disk and bulge
back to the gas particle from which it formed. Each gas particle is tagged with a local
density and temperature that is a function of the simulation force resolution and softening
length. Gas properties are calculated based on the 32 nearest neighbors. Our definition
of pressure is limited to a “thermal” pressure term, which, as we do not resolve disks of
highly turbulent gas, is actually a proxy for the entire pressure in the gas. Namely, it is the
primary pressure support against the gravitational pressure in the disk.
3.3
Results
In this section we show that ISM structure is closely tied to star formation. Additionally,
we show that at earlier times in a galaxy’s history, stars are forming in a higher pressure
ISM environment than that in which stars form today. This is theoretical evidence for a
variation in IMF in a “normal” Milky-Way environment, if IMF variations are indeed tied
to ISM structural parameters, like temperature, density, metallicity and pressure (Conroy
& van Dokkum, 2012; Krumholz, 2011). For brevity, we show plots for one of our simulated
halos, demonstrating the trends observed for all four halos.
In all four simulated halos, each cosmological and with varying merger and star formation histories, the distribution of pressure values in the ISM that forms the present day bulge
is higher by an order of magnitude than that which forms the present day disk. In Figure
3.1a, we show the distribution of formation pressures for one of our simulated Milky-Ways,
h986. The peak pressure of both distributions is different: bulge stars peak at pressures
an order of magnitude higher pressure than disk stars. This shows that in general, bulge
54
stars are forming ISM that is structurally different in terms of gas temperature and density:
specifically, stars are forming from denser gas. It is important to note that the actual values
for pressure are not comparable to pressures found in observations. As our star formation
prescription is resolution limited, the maximum gas densities achieved are resolution dependent. What should be highlighted is the relative difference between the pressures found in
the bulge and disk of our simulated galaxies. In Figure 3.1b and 3.1c, we show the SFHs
for the 3 simulated halos, not directly discussed here, to show that in general, for all halos,
the bulge forms at early times, and the disk forms later.
In Figure 3.2 we compare the phase diagrams for bulge and disk during a star formation
event which contributes to the components’ overall mass growth: for the bulge, this was
between 2.5 and 4 Gyrs and for the disk, between 10 and 13 Gyrs. This figure at first
glance, demonstrates that in general, the the bulge forms in a range of densities that is
higher than that of the disk, and that the temperature range is very similar. Each point
in the phase diagrams is color coded according to pressure where hotter colors represent
higher pressures (red is the highest pressure bin). This color-coding further drives home
that it is the high densities in the bulge star formation event that drive the high pressures,
while in the disk, the high pressures result from higher temperatures. Since our simulations
use an H2 -dependent star formation recipe, low metallicity gas particles would be expected
to form stars at higher densities. However, from the bottom panel of Figure 3.2 is is clear
that even at the same metallicities, bulge stars form from denser gas.
In Figure 3.3, we present the star formation history for the same galaxy, h986, for
the dynamical bulge (red) and disk (blue). We also present the median pressure as a
function of time for both components in the same time bins. Figure 3.3 highlights that
both components are forming stars at higher pressures early in the galaxy’s history. The
star formation histories show that the bulge forms the majority of its stars early on, when
typical ISM pressures are high, while the disk forms its stars later, when ISM pressures are
lower. We also can see the parallel in the bulge SFH and the bulge pressure, in that bursts
of bulge star formation seem to be contemporaneous with ISM pressure peaks. We discuss
55
what this may imply in the summary.
Finally, in Figure 3.4, we explore whether formation location has any bearing on the
pressure of the gas. We expect that pressure is higher closer to the center of the galaxy
(i.e., where one would expect to find bulge stars), given that the vertical gravity and surface
density should be higher closer to the center. As a result, one would expect high pressures
towards the center of the galaxy. However, Figure 3 shows that this explanation cannot
entirely explain the pressure differential between bulge and disk. In the top panel we see that
there is no correlation between formation radius and pressure for either component, over
the galaxy’s whole history and that bulge stars are in forming at higher pressures than disk
stars. In the second panel, we look only at the stars that formed in the early protogalaxy:
specifically, the gas that forms stars that are 6 Gyrs old or more, when both components
are should be at higher pressures. We see that with that cut, at any formation radius,
bulge stars and disk stars are forming at high pressures in the early universe, implying the
existence of an early high pressure star-forming environment in the protogalaxy.
3.4
Summary
In this letter, we provide evidence that ISM pressure is redshift dependent by examining
the ISM pressures during the formation of the present-day bulge and disk. Because the
present-day bulge predominantly forms early in the galaxy’s history, it forms at higher
pressures than the present-day disk. We show that in general, at early times, star formation
occurs at higher pressures- specifically higher densities. We show that this is not the result
of formation location and the higher densities found in the center of the protogalaxy: in
general bulge and disk stars are forming over all formation radii. If ISM pressure and
IMF are related as postulated in Krumholz (2011); Conroy & van Dokkum (2012), we have
evidence for a redshift dependence of the IMF and further, that bulge stars formed with a
different IMF from disk stars. Furthermore, we show that even at high redshift, bulge stars
and disk stars in general, are forming at higher pressures, regardless of formation radius.
This further supports the redshift dependence of ISM pressure: we see that in the young
56
protogalaxy stars form in a high pressure disk, regardless of classification at z = 0. However,
as Figure 3.3 demonstrates, the majority of bulge stars are formed in this high-pressure star
formation epoch, while the majority of disk stars form in the later, low-pressure epoch. In
our analysis we also examine the differences in metallically and H2 fraction between bulge
and disk, to isolate why the early protogalaxy is at high density (and thus high pressure).
We find that bulge stars form from gas with higher H2 fractions and even when holding
metallicity constant, bulge stars form from denser gas. These trends point toward the high
redshift progenitors having more dense gas, likely as a result of early rapid accretion.
Future work will include following the assembly history of these galaxies to determine
the role of mergers and gas accretion in the formation of the present-day bulge and disk.
We will follow the build-up of each component tracing each star back in time, including a
full merger-tree. In this way we can determine isolate the role of mergers and in-situ star
formation on the structure of the star-forming ISM
3.5
Supplementary Analysis
In this section I discuss supplementary figures that are not included in the analysis submitted
to ApJL above. In the analysis below, I try to figure out why the pressure of the star forming
ISM is time variant, resulting in the difference in pressure between bulges and disks in our
simulations. In Figure 3.5, I examine the distribution of metallicities between bulge and disk
to explore whether metallicity plays a role in the increased density that results in higher
pressure at early times. However, metallicity peak for both the bulge and disk are very
similar, although the disk component does have a high metallicity tail.
In Figure 3.5, I examine the distribution of H2 mass fractions in the bulge and the disk
and find that there is not a significant difference betwen the mass fractions in the bulge
and disk of the galaxy, although the bulge, in general, has a slightly lower average H2 mass
fraction. I look at this more closely, by looking at the evolution as a function of time of the
H2 mass fraction over the course of a star formation event in each component (i.e. when
that component was actively forming stars) in Figure 3.6. We see that on average, the disk
57
has a lower average mass fraction over the course of it’s burst when compared to that of
the bulge.
Finally in Figure 3.7, I examine where on the phase diagram different values of H2 mass
fraction occur in the bulge: blue for high mass fractions, and red for low. In this plot it
is interesting to note that all the high H2 mass fractions correspond to the high densities
in the phase diagram, meaning that there may be a correlation between an enhanced mass
fraction and enhanced pressure.
58
Figure 3.1 Top Panel: Distribution of pressures for bulge and disk in one of the simulated
galaxies, h986. Note that the peak of the distribution of pressures of the bulge is higher
than the peak of the pressure distribution of disk stars. Middle Panel: SFHs for the bulges
of the 3 galaxies not shown in this manuscript. Note that like h986 shown in Figure 3, the
bulges form early in the galaxy’s history. Bottom Panel: SFHs for the disks of the 3 galaxies
not shown in this manuscript. As with the bulges, these galaxies also follow the same trend
as h986, with disk star formation occuring later in the galaxy’s history.
59
600
T (K)
400
200
0
-5
-4
-3
log(nH (cm-3))
-2
-1
-4
-3
log(nH (cm-3))
-2
-1
600
T (K)
400
200
0
-5
Figure 3.2 Phase Diagrams for bulge (top) and disk (bottom) during a star formation event,
color coded by pressure. Star formation events for each component were selected based on
contribution to each components’ overall growth. Hotter colors are higher pressures, cooler
colors are lower pressures. Note that high pressures are driven by high densities in the
bulge. In the bottom panel we show the metallicity of the gas that formed both the bulge
and disk stars versus its density at the time of star formation (note that both density and
metallicity are smoothed over hundreds of parsecs).
60
Figure 3.3 Top Panel: Star formation rates for each of the dynamical components of h986.
Bottom Panel:Pressure vs. Time for each of the components. This highlights the redshift
dependence of the pressure of the star forming ISM: early on, stars are forming at higher
pressures, regardless of which component they belong to at z = 0. Note also the peaks in
pressure are present when the SFH is peaking, in bulge stars. The big bursts in the bulge
SFH correspond to major mergers in the galaxy’s history.
61
Figure 3.4 Top Panel: Pressure vs. formation radius for the bulge and disk, over the galaxy’s
whole history. Bottom Panel:Pressure vs.formation radius for each of the components, for
stars that formed before 6 Gyrs (when the pressure of the ISM was higher for both components. This figure highlights that formation radius is not the underlying cause of the
pressure differential between bulge and disk and that on average, bulge stars are forming
at higher pressures than disk stars. The bottom panel shows that in the first half of the
galaxy’s history, stars are forming at higher pressures in general, regardless of component
and formation radius.
62
1.0
Disk Stars
Bulge Stars
0.8
0.6
0.4
0.2
0.0
0.05
0.10
0.15
0.20
metalicity (z/zsol)
0.25
Figure 3.5 Distribution of metallicities for bulge and disk in one of the simulated galaxies.
Note that the peak of the distribution of pressures of the bulge and disk are nearly the
same.
63
1.0
Disk Stars
Bulge Stars
0.8
0.6
0.4
0.2
0.0
-6
-4
-2
Log(H2 mass frac)
0
Figure 3.6 Distribution of H2 mass fractions for bulge and disk in one of the simulated
galaxies. Note that the peak of the distribution of pressures of the bulge and disk are nearly
the same.
64
Figure 3.7 Average H2 mass fractions for the bulge as a function of time, during a star
formation event.
65
Figure 3.8 Average H2 mass fractions for the disk as a function of time, during a star
formation event.
66
Figure 3.9 Phase diagram for bulge stars color coded by H2 fraction. Blue points are high
H2 fractions, and red points are low H2 fractions.
67
Chapter 4
DWARFS AND STAR FORMATION (SF) INDICATORS: ARE SF
INDICATORS SENSITIVE TO TIMESCALE?
In this chapter, I discuss early work-in-progress results on dwarfs and star formation
indicators. I use observational techniques to measure star formation rates in dwarf galaxies,
using simulated observations generated from the postprocessing software, SUNRISE. We
show that our cosmological dwarf galaxy simulations have star formation rates similar to
those observed locally, in the LVL sample (observational data courtesy of Weisz et al.
(2012)), regardless of star formation recipe and resolution. Further, we show preliminary
analysis of the timescale dependence of SFR on the indicator used.
4.1
Introduction
Observations of star formation rates (SFRs) of galaxies provide vital clues into the history
of the galaxy and are key probes into the evolutionary history of the galaxy. Different
types of galaxies along the hubble sequence contain a wide range of young stellar content
and star formation activity. Understanding this variety and its origin is fundamental to
understanding galaxy evolution in the context of it’s stellar populations.
Hα nebular line emission and the UV continuum are two widely used star formation
rate (SFR) indicators. The Hα emission traces the recombination of the gas ionized by
the most massive stellar population, and therefore traces star formation over the lifetime of
these massive O and B stars (masses greater than 20 solar masses). UV flux, on the other
hand, originates from the photospheres of stars from O through later B type stars (any stars
greater than 3 solar masses) and thus measures a longer timescale: averaged over nearly
108 years. Both tracers should yield consistent SFRs as both are tracing the young stellar
population in galaxies. Discrepencies between the two measures reveal problems either in
68
the calibration of these indicators, or more fundamentally, the assumptions in converting
these fluxes into star formation rates. Star formation rates from indicators are generally
derived assuming this functional form: SFR = K(λ)×F(λ), where K(λ) can be inferred
from population synthesis models under several assumptions, including a functional form of
the IMF. This assumes a universal IMF, regardless of morphological type, luminosity and
redshift. Additionally, it assumes certain timescales of star formation: for example it has
been shown that a bursty star formation history, like that expected in dwarfs, can cause
variations in the indicators, such that they do not trace each other (Mateo, 1998).
Many studies have directly compared the integrated Hα and UV SFRs for which both
diagnostics have been measured (Buat et al., 1987; Buat, 1992; Bell & Kennicutt, 2001; Salim
et al., 2007). Normal spiral galaxies show broad agreement between the two indicators, with
the correct treatment of dust attentuation and extinction (they find that UV is the most
impacted by extinction). As lower mass regimes are probed, as mentioned above, there are
indications that the flux ratio systematically decreases and that Hα flux is systematically
lower when compared to the UV (Bell & Kennicutt, 2001). This trend could mean many
things: recent star formation histories of these galaxies are bursty or abrupt and it could
also mean, the IMF is not universal and becomes more deficient in the highest mass stars
in these low mass, low surface brightness systems.
In this chapter, I will address how I used simulated dwarf galaxies in an attempt to
track what is going on with these star formation indicators. I first discuss the simulations
that I used in this analysis and the affects of resolution. Then I discuss the software I used
to create the artificial observations used for making measurements. I show the agreement
between the simulations and the observed trends with mass.
4.2
The Simulations and Analysis
The production-run simulations used in this chapter were, as were all the other simulations,
run with the N-Body + SPH code GASOLINE (Wadsley et al., 2004; Stinson et al., 2006)
in a fully cosmological ΛCDM context: Ω0 = 0.26, Λ=0.74, h = 0.73, σ8 =0.77, n=0.96.
69
This set of simulations includes metal line cooling (Shen et al., 2010) and non-equilibrium
model for H2 abundance, which includes formation on dust grains, destruction by LymanWerner radiation and shielding of both HI and H2 (Christensen et al., 2012d), as described
in detail in chapter 2.
The older, lower resolution simulations used in this chapter were also run with GASOLINE with the same cosmology, and like the production-run simulations, selected from
uniform DM-only simulations of 25 Mpc per side. The same field–like regions were selected
and resimulated at higher resolution, but these were resimulated at half the effective resolution of the new simulations described above. These simulations incorporated detailed
metal cooling and turbulent mixing in the ISM. There is a metallicity dependent radiative
cooling included at all the temperatures in the range of 100-109 K. The cooling function is
determined using pre-computed tabulated rates from the photoionization code CLOUDY
(Ferland et al., 1998a), folowing Shen et al. (2010). These values assume that metals are in
ionization equilibrium and the ionization, cooling and heating rates for primordial species
are calculated time-dependently from the rate equations.
Both sets of simulations utilized the “blastwave” SN feedback approach. As in previous
works using the “blastwave” SN feedback approach (Stinson et al., 2006; Governato et al.,
2012), mass, thermal energy, and metals are deposited into nearby gas when massive stars
evolve into SNe. The amount of energy deposited amongst those neighbors is 1051 ergs per
SN event. Gas cooling is then turned off until the end of the snow plow phase of the SN
blastwave. Equilibrium energy rates are computed from the photoionization code Cloudy
(Ferland et al., 1998b), following Shen et al. (2010). A spatially uniform, time evolving,
cosmic UV background turns on at z = 9 and modifies the ionization and excitation state
of the gas, following an updated model of Haardt & Madau (1996). The simulations also
include a scheme for turbulent mixing that redistributes heavy elements among gas particles
(Shen et al., 2010).
Interfacing GASOLINE with SUNRISE (Jonsson et al., 2010; Jonsson, 2006) allows us
to measure the SED of every resolution element of our simulated galaxies from the far UV
70
to the far IR with a full 3D treatment of radiative transfer. SUNRISE includes a realistic
description of the effects of dust reprocessing: an important source of bias in observations
(Narayan 2010). The latest version enables us to make detailed calculations of images,
spectra and integral field unit-type outputs of line-of-sight velocity distributions, accurately
accounting for the scattering and differential attenuation of light from different populations
along the line of sight. In this analysis, we utilize the extensive library of filters ranging
from GALEX FUV through Spitzer MIPS, as well as create our own filter, for the Hα line
to compare the broadband flux with the lineflux from the SUNRISE generated SED.
In Figure 4.1, we show the relevant portion of the SUNRISE SED, showing the Hα line
from which we measure the lineflux to derive a SFR. In Figure 4.2, we show a mock-image
generated by SUNRISE, in the FUV, of one of our dwarf galaxies. In Figure 4.3, we show the
same mock image convolved with GALEX’s PSF, which creates the final image from which
we make our measurement. Our FUV measurements of flux follow the flux measurement
techniques utilized by the observers when making their measurements (i.e., we perform
our photometry using the methodology outlined on the GALEX website). We obtain Hα
lineflux measurements by measuring the continuum of the artificial SED for the 100-300
angstroms adjacent to the line. We then measure the flux of the line for various widths,
based on fitting a gaussian line profile to the line to get a baseline width. I also developed
code to create artificial Hα images using SUNRISE by creating a narrow-line filter to add
to the SUNRISE library.
4.3
Results
In this section we show the mock images generated by SUNRISE, as well as the relevant
portions of the SED utilized in this analysis. We also show that measurements of star
formation rate based on these mock images and SEDs result in good agreement between
simulations and observations. In figure 4.4, we show the relationship between the star
formation rate derived from UV to that derived utilizing the Hα lineflux from the SED
using our current production-run simulations that include H2 based star formation, and
71
1.5•1043
Flux (W/m)
1.0•1043
5.0•1042
0
6.0•10−7
6.2•10−7
6.4•10−7
6.6•10−7
Wavelength(m)
6.8•10−7
7.0•10−7
Figure 4.1 Simulated observation of Hα emission from SUNRSISE SED. The lineflux from
this line is used to derive SFR in simulated dwarfs, normalized by continuum measurements
in the simulated SED.
double the resolution. We show that the simulated galaxies follow the general trend of the
observed galaxies (black points), in the mass range probed by our simulations.
In Figure
4.5 we show the ratio of SFRs derived in the low resolution, metal-line cooling simulations.
For some of the galaxies, we also show the z=0.5 ratio to see if there is a time dependence to
the agreement between simulations and observations. Again, even with the different cooling
model, and lower resolution, the simulations show broad agreement with the trends shown
in the observational data.
In Figure 4.6 we show a comparison between a star formation history for the same dwarf
in the low resolution metal-line cooling run, and the high resolution H2 run. We note that
the former forms more stars over the galaxy’s history and has in general, what appears to be
a more bursty star formation history (SFH). In the newer run, there is less star formation
and the SFH is less punctuated with big bursts at early times. However, we note that in
72
4
kpc
2
0
-2
-4
-4
-2
0
kpc
2
4
M arcsec-2
30 29 28 27 26 25 24 23 22 21 20 19
Figure 4.2 Simulated observation of 24 micron emission from SUNRISE. The flux from
thses images is used to derive SFR in simulated dwarfs.
the last 4 Gyrs, both galaxies have a quiescent, relatively constant SFR. Both runs have
SFRs derived from the indicators that are consistent (0.03Msun yr−1 ). As the star formation
indictors are tuned such that they are sensitive to recent star formation, we expect there to
be agreement between the runs, as they differ mainly at early times in their star formation
history. This regime is where the low resolution run produces more stars, which should not
73
4
kpc
2
0
-2
-4
-4
-2
0
kpc
2
4
M arcsec-2
30 29 28 27 26 25 24 23 22 21 20 19
Figure 4.3 Simulated observation of UV emission from SUNRSISE generated image, including dust reprocessing. The flux from such images are utilized to derive SFRs.
change measurements derived from indicators probing young populations.
In Figures 4.7 and 4.8, we examine measured SFR from an indicator compared to the
actual SFR from the simulation, integrated over various timescales. We expect that if a
given indicator is most sensitive over a specific timescale, the recovery of the true SFR will
be closer to 1. However both these plots for the UV and Hα are inconclusive. Niether
74
log[SFR(Hα)/SFR(UV)]
1
0
-1
-2
6
7
8
9
log(Mass)MΟ•
10
11
Figure 4.4 Ratio of SFRs derived from the UV continuum and from Hα. Stellar mass is
on the x-axis, and the ratio of SFRs from the various indicators are on the y-axis. Green
squares are the simulations, and the black points are the data from Weisz et al. (2012).
indicator seems to recover SFR well, no matter the timescale. Additionally, the time trends
are not what is expected- we are seeing good recovery at 1 Gyr, although neither indicator
should have better sensitivity on such a long timescale. We expect that this may be a mass
effect: these halos are too massive to distinguish the mass effect as we’d expect stronger
trends in lower mass halos. In smaller halos feedback shuts off the star formation entirely
for some period of time. As a result, Hα shuts off very quickly, and the UV, sensitive to
less massive stars, will linger a bit longer. However, in these more massive halos, there is
always a low level of SF resulting in there always being both Hα emission and UV flux.
75
Figure 4.5 Ratio of SFRs derived from the UV continuum and from Hα, utilizing lower
resolution simulations, with only metal-line cooling. Each color represents a different dwarf
simulation; if more that one point exists for a given color, one point is z=0 and the other
is at z=0.5. The greyed out points in the backgroud once again are the observational data
from Weisz et al. (2012).
4.4
Summary
In this chapter, I show that there is broad agreement between SFRs derived from simulated
dwarf galaxies, and those derived from observations. I show that even with changing the
star formation model and the resolution of the simulations, we still show agreement between
simulations and observations. Further study would include tracing the time-dependence of
these indicators over the course of a star formation burst to see how sensitve the star formation indicators are to the burstiness, and when during a burst, you are probing the galaxy.
76
My work in progress includes running an extremely high resolution sheet of dwarf galaxies,
using the newest version of GASOLINE, ChaNGa, which fixes the problems associated with
traditional SPH simulations. Using this new sample of dwarfs, I will be able to probe a lower
mass regime and study the time scale dependence of the indicators in greater detail over
a lower mass range that has greater inconsistency between indicators in the observational
data.
77
Figure 4.6 Top Panel: Star formation rate for the lower resolution, metal-line cooling run.
Bottom Panel:Star formation rate for the high resolution, H2 star formation run.
78
1.0
Measured SFR/Actual SFR
0.8
0.6
0.4
0.2
0.0
5
50
1000
Timescale (Myrs)
Figure 4.7 Measured Hα SFRs versus actual SFRs integrated over various timescales. Each
color represents a different halo. Hα SFRs measured using the lineflux from the simulated
SED. The measured SFRs come directly from the simulations, integrated over the timescale
shown on the x-axis. This plot, although inconclusive, shows that the SFR does change as
a function of timescale it’s integrated over, although no single timescale appears to match
with the Hα SFRs in this mass range.
79
1.0
Measured SFR/Actual SFR
0.8
0.6
0.4
0.2
0.0
5
50
1000
Timescale (Myrs)
Figure 4.8 Measured UV continuum SFRs versus actual SFRs integrated over various
timescales. Each color represents a different halo. Hα SFRs measured using the lineflux
from the simulated SED. The measured SFRs come directly from the simulations, integrated
over the timescale shown on the x-axis. This plot, although inconclusive, shows that the
SFR does change as a function of timescale it’s integrated over, although no single timescale
appears to match with the UV SFRs in this mass range.
80
Chapter 5
CONCLUSIONS AND FUTURE WORK
5.1
Conclusions
This thesis broadly encompasses the complex topic of star formation, specifically in Nbody+SPH cosmological simulations. Specifically this thesis is a step toward answering the
following questions:
How do galaxies populate their dark matter halos? What affect does feedback
and baryonic physics have on the stellar to halo mass relationship?
We measured the SHM (stellar mass – halo mass) relation for a set of simulated field
galaxies and compared it with the redshift zero predictions based on data from the SDSS
and the Abundance Matching Technique described in M12. We find very good agreement in
normalization and shape over five orders of magnitude in stellar mass with the abundance
matching result from M12. This agreement between simulations and observational data is
due to two systematic factors: 1) An implementation of SF that relates the SF efficiency
to the local H2 abundance in resolved star forming regions, resulting in localized feedback
that significantly lowers the SF efficiency and 2) “observing” the simulations to properly
compare them to observational estimates of the SHM relation. Our analysis shows that
adopting photometric stellar masses contributes to a 20-30% systematic reduction in the
estimated stellar masses. Stellar mass estimates based on one band photometric magnitudes are likely to underestimate the contribution of old stellar populations (reflecting the
larger contribution to the total flux coming from younger stars). This systematic effect is
further exacerbated by the use of aperture based magnitudes, adding another 20-30% due
to neglecting the contribution of low surface brightness populations. Finally, a third systematic effect comes from a difference in halo masses in collisionless (DM-only) simulations
81
vs simulations including baryon physics and outflows. Baryon mass loss makes halo masses
smaller by up to 30% when calculated at the same overdensity.
When and where are stars forming in galaxies? What is the structure of
the ISM where these stars are forming? How does this relate to the underlying
IMF of these systems?
We provide evidence that ISM pressure is redshift dependent by examining the ISM
pressures during the formation of the present-day bulge and disk in 4 simulated disk galaxies.
Because the present-day bulge predominantly forms early in the galaxy’s history, it forms
at higher pressures than the present-day disk. We show that in general, at early times,
star formation occurs at higher pressures- specifically higher densities. We show that this
is not the result of formation location and the higher densities found in the center of the
protogalaxy: in general bulge and disk stars are forming over all formation radii. If ISM
pressure and IMF are related as postulated in Krumholz (2011); Conroy & van Dokkum
(2012), we have evidence for a redshift dependence of the IMF and further, that bulge stars
formed with a different IMF from disk stars.
How well do observational star formation indicators trace star formation?
What biases do we see when comparing simulations to observations? What
does this tell us about the IMF in dwarf galaxies?
I show that our simulated galaxies have good agreement with the observed star formation
rates derived from Hα and the FUV. I show that this agreement holds over an order of
magnitude in resolution, and over two different ISM cooling subgrid models. I also looked
at the variation in the measurement between z=0 and z=0.5 and find that over that timescale
there is still good agreement between simulations and observations. In this section of my
thesis, I showed the efficacy of utilizing SUNRISE to generate mock observations (images
and SED) and utilizing these images to make measurements on the simulated galaxies. Now
that the tools for implementing these measurements are built, I can answer questions about
the variations in SFR derived from the two indicators discussed and trace their variation
as a function of time over the course of a star formation burst. This ideally will address
82
the question of what causes the variation: whether it soley is due to the bursty nature of
dwarf star formation histories, or if the high mass end of the IMF is in fact different in these
galaxies.
5.2
Future Work
Recent advances in imaging capabilities are now providing exquisite data on the study
of high-redshift progenitors of today’s massive galaxies. These high redshift galaxies are
currently providing us with insights into the key mechanisms that shape the evolution of
galaxies in the high mass regimes, placing constraints on current models of galaxy formation and evolution. In particular, the new near-infrared HST WFC3/IR imaging and
spectroscopy provided by the Cosmic Assembly Near-IR Deep Extragalactic Legacy Survey
(CANDELS) (Grogin et al., 2011; Koekemoer et al., 2011) and 3D-HST (van Dokkum et al.,
2011), provide the necessary combination of depth, angular resolution and area to enable
the most detailed and robust studies to date of rest-frame optical morphologies of galaxies
at redshifts of 1 < z < 3. In future work, I would like to run a series of cosmological hydrodynamic simulations at unprecedented high-resolution, that will resolve, down to 50 pc (2x
better than the leading current simulations) scales, the ISM structure, star formation and
physical structure of massive spirals. These simulations are key for the interpretation of
recent observational results from CANDELS and 3D-HST that challenge the simple mergerdriven picture of galaxy growth and evolution currently in favor. Our simulations already
are able to produce galaxies with realistic present-day properties: including but not limited
to morphology, stellar masses and baryon fractions (Munshi et al., 2013b; Governato et al.,
2009, 2010). This new, state of the art suite of simulations will answer the following key
questions:
• When and how do bulges assemble, acquire their mass and evolve in size?
• Given the present day morphology of a massive disk galaxy, when and where did the
stars form in each component, as a function of time?
83
• How are star formation, assembly history and ISM structure related in massive disk
galaxies?
In short, the implemenation and analysis of the simulations will provide critical observationdriven constraints to theoretical models of bulge and disk growth including the transition
between bulge and disk dominated systems, and explain the relationship between star formation in progenitors of massive bulges and the underlying ISM and IMF.
HST high resolution data from WFC3 can provide the detailed morphology of high–z
massive galaxies (Schawinski et al., 2011; Bundy et al., 2005). These high–z data have
resolution comparable to or better than the local data from the Sloan Digital Sky Survey
(SDSS). My future work will help interpret results from CANDELS and 3-D HST, which are
providing a detailed census of massive galaxies at redshifts 1 < z < 3. 3-D HST is a nearinfrared spectroscopic survey designed to study the physical processes that shape galaxies
in the distant universe. This survey provides rest-frame optical spectra for essentially every
object in the field and is optimized to study galaxy formation within 1 < z < 3. Its
science objectives specifically address resolving the growth of disks and bulges, spacially
and spectrally in addition to disentangling the processes that regulate star formation in
massive galaxies. CANDELS, a complement to 3D-HST, uses HST to obtain deep images
of more than 250, 000 galaxies with WFC3/IR and ACS imaging, over many fields imaged
with other instruments. One of its key objectives is to study the “cosmic noon” and use
these rest-frame optical observations to provide solid estimates of bulge and disk growth.
Clearly, one of the key goals of both the CANDELS and 3D-HST surveys includes using
rest-frame optical observations and spectroscopy at 1 < z < 3 to provide solid estimates of
bulge and disk growth and formation. Both surveys allow astronomers, for the first time to
directly study high redshift morphologies of galaxies.
Published data from these surveys, empirically explore the evolutionary factors that lead
to a lack of star formation in massive galaxies and find evidence for dramatic changes in the
morphologies of massive galaxies with redshift, specifically at z ∼ 2, marking a key transition
from being disk to bulge dominated (Bell et al., 2012; Bruce et al., 2012). Additionally, while
84
most of the quiescent galaxies are bulge-dominated, the data indicate that a significant
fraction of quiescent galaxies in this redshift range, still have disk-dominated morphologies.
Furthermore, these data highlight that bulges are more compact with increasing z, although
there is no significant mass evolution (Bruce et al. 2012).
Additionally, recent observational and theoretical work (Conroy & van Dokkum, 2012;
Krumholz, 2011) indicate that in spheroidal, high mass systems (i.e., high velocity dispersion
systems), stars are forming in a higher pressure ISM when compared to stars formed in
other systems. The higher pressure ISM correlates with the initial mass function (IMF)in fact, higher pressures indicate a more bottom heavy IMF. It is suggested in Conroy and
VanDokkum (2012) that this IMF variation is correlated, not only with the pressure of
formation, but additionally with the intensity of star formation: higher SFR densities and
higher pressures correspond to a more bottom heavy IMF. This project will provide critical
theoretical constraints to the above results from the CANDELS and 3D-HST surveys by
directly tracing the formation of bulge and disk components of massive galaxies in high
resolution simulations. As a result, this work will directly address the observers’ key goal of
understanding bulge and disk growth and subsequent evolution, in massive spirals.
The above observational evidence challenges theoretical models to: (1) include a mode
in which star formation quenching is not connected to morphological transformation, (2)
explain the relationship between active and passive disks and, (3) predict the rapid demise
of star forming disks and the slow emergence of bulge-dominated morphologies. Theoretical
evidence points out that the simple therory of merger driven assembly of bulges (Hopkins
et al., 2005; Oser et al., 2012), is likely incomplete and and too simplistic. Rather, HST
data point to an early growth of a compact bulge and subsequent size evolution in later
epochs. The high-z evidence is in apparent contrast with the low-z size-mass relationship for
galaxy spheroids suggesting that secular processes outside of merging may be responsible
for morphology evolution.
85
As stated above, I will focus on very high resolution ’zoomed-in’ simulations in a ΛCDM
Universe. These simulations will naturally include mergers, feedback, cold flows and cooling
of shock heated gas (Kereš et al., 2009; Brooks et al., 2009; Dekel et al., 2009) in a cosmological setting and will provide a detailed approach to the growth of spheroids and disks
over the whole Hubble time.
ChaNGa: An NBody+SPH code with star formation linked to H2 and SN
and SMBH Feedback.
GASOLINE, our parallel SPH code, has received two major updates: Collaborator Wadsley has introduced a new SPH implementation (Hopkins, 2013) that successfully models
Kelvin-Helmholtz instabilities and passes all the tests outlined in Agertz et al. (2007). This
makes upgraded GASOLINE, now called ChaNGa, competitive with the new proposed
grid codes, while keeping the versatility and simplicity of its Lagrangian approach. Quinn
and collaborators have adapted ChaNGa to an advanced parallel run-time system with a
dynamic load balancing scheme (CHARM++) to scale efficiently up to housands of CPUs
and enable efficient use of heterogeneous architectures. Advantages of this new code include:
• a new SPH treatment of hydrodynamics and diffusion processes
• Scaling up to several thousands of CPUs- allowing simulations to run faster and more
efficiently.
• Star Formation with different IMFs, O and Fe metal enrichment.
• Gas heating from SN feedback, SMBHs and a cosmic UV background including QSOs.
• H2 shielding and metal line cooling; SF related to the local abundance of H2 .
How can numerical simulations address the above questions?
The key questions I hope to address can now be directly tackled by numerical simulations
in a full cosmological setting. Our understanding of the formation of galaxy bulges and
disks and the quality of hydrodynamical methods has greatly improved in the past few
years, since simulations are now able to resolve scales ∼ 5% of the typical disk scale length
and conserve the angular momentum of infalling gas (Robertson et al., 2004; Brooks et al.,
86
2011; Keres et al., 2011). Recent work has highlighted the role of feedback (from SN and
BHs) and has demonstrated that galaxy outflows are efficient at removing baryons and
possibly self–regulate the growth of galaxies and their stellar masses (Arav et al., 2008;
Bouché et al., 2012; Governato et al., 2010; Di Matteo et al., 2005; Oppenheimer et al.,
2010; Choi & Nagamine, 2011; Munshi et al., 2013b). Simulations coupled with analytical
modeling of processes at unresolved scales can now achieve the dynamical range and physical
detail necessary to follow the assembly of massive disk galaxies and the their corresponding
components over a Hubble time. These high resolution simulations can now reliably predict
morphological evolution and take CANDELS observations as model constraints.
Because of this vast improvement in hydrodynamical models, I can use my proposed
simulations to directly address the proposal’s key questions:
When and how do bulges assemble, acquire their mass and evolve in size?
Given the present day morphology of a massive disk galaxy, when and where
did the stars form in each component, as a function of time?:
Using a dynamical decomposition based on energy and angular momentum, I can decompose each of the simulated galaxies into its structural components: bulge, disk and halo.
I will do this from z = 0 to z = 3 and follow where the stars that form each structural
component come from, as a function of time, in order to understand the assembly history
of the bulge, including utilizing full merger trees for each of the simulations.
How are star formation, assembly history and ISM structure related in massive disk galaxies?: Is there a connection between ISM pressure, merger history, and
assembly history and the underlying IMF in the components of these galaxies (Krumholz,
2011; Conroy & van Dokkum, 2012)? My results (Munshi et al., 2013a) indicate that the
structure of the ISM when the bulge stars formed is different than that of the ISM when disk
stars formed, possibly supporting observational evidence for a varying IMF. What remains
to be answered is what actually causes this variation: what in the history of these galaxies
leads to this pressure differential? I have already developped the software to trace stars back
87
to the ISM in which they formed and can apply this software to the new, much more massive
simulations. Coupled with the analysis of how each component forms, in addition to generating full merger trees, I can trace how and why the star forming ISM varies, what causes
the variation in ISM pressures and potentially isolate which process drives a time-dependent
IMF.
88
BIBLIOGRAPHY
Abadi, M. G., Navarro, J. F., Steinmetz, M., & Eke, V. R. 2003, ApJ, 591, 499
Agertz, O. et al. 2007, MNRAS, 380, 963
Arav, N., Moe, M., Costantini, E., Korista, K. T., Benn, C., & Ellison, S. 2008, ApJ, 681,
954
Avila-Reese, V., Colı́n, P., González-Samaniego, A., Valenzuela, O., Firmani, C., Velázquez,
H., & Ceverino, D. 2011, ApJ, 736, 134
Barnes, J., & Efstathiou, G. 1987, ApJ, 319, 575
Bastian, N., Covey, K. R., & Meyer, M. R. 2010, ARA&A, 48, 339
Behroozi, P. S., Conroy, C., & Wechsler, R. H. 2010, ApJ, 717, 379
Behroozi, P. S., Wechsler, R. H., & Conroy, C. 2013, ApJ, 770, 57
Bell, E. F., & de Jong, R. S. 2001, ApJ, 550, 212
Bell, E. F., & Kennicutt, Jr., R. C. 2001, ApJ, 548, 681
Bell, E. F., McIntosh, D. H., Katz, N., & Weinberg, M. D. 2003, ApJS, 149, 289
Bell, E. F. et al. 2012, ApJ, 753, 167
Benson, A. J. 2010, 495, 33
Benson, A. J., Frenk, C. S., Lacey, C. G., Baugh, C. M., & Cole, S. 2002, MNRAS, 333,
177
Bigiel, F., Leroy, A., Walter, F., Brinks, E., de Blok, W. J. G., Madore, B., & Thornley,
M. D. 2008, AJ, 136, 2846
89
Binney, J., Gerhard, O., & Silk, J. 2001, MNRAS, 321, 471
Blain, A. W., Chapman, S. C., Smail, I., & Ivison, R. 2004, ApJ, 611, 725
Blanton, M. R. et al. 2001, AJ, 121, 2358
Blumenthal, G. R., Faber, S. M., Primack, J. R., & Rees, M. J. 1984, Nature, 311, 517
Bode, P., Ostriker, J. P., & Turok, N. 2001, ApJ, 556, 93
Bouché, N., Hohensee, W., Vargas, R., Kacprzak, G. G., Martin, C. L., Cooke, J., &
Churchill, C. W. 2012, MNRAS, 426, 801
Bower, R. G., Benson, A. J., & Crain, R. A. 2012, MNRAS, 422, 2816
Bower, R. G., Benson, A. J., Malbon, R., Helly, J. C., Frenk, C. S., Baugh, C. M., Cole,
S., & Lacey, C. G. 2006, MNRAS, 370, 645
Bower, R. G., Vernon, I., Goldstein, M., Benson, A. J., Lacey, C. G., Baugh, C. M., Cole,
S., & Frenk, C. S. 2010, MNRAS, 407, 2017
Boylan-Kolchin, M., Bullock, J. S., & Kaplinghat, M. 2011a, ArXiv e-prints
—. 2011b, MNRAS, 415, L40
Boylan-Kolchin, M., Springel, V., White, S. D. M., Jenkins, A., & Lemson, G. 2009, MNRAS, 398, 1150
Brook, C. B. et al. 2011, MNRAS, 415, 1051
Brook, C. B., Stinson, G., Gibson, B. K., Wadsley, J., & Quinn, T. 2012, MNRAS, 424,
1275
Brooks, A. M., Governato, F., Booth, C. M., Willman, B., Gardner, J. P., Wadsley, J.,
Stinson, G., & Quinn, T. 2007, ApJ, 655, L17
Brooks, A. M., Governato, F., Quinn, T., Brook, C. B., & Wadsley, J. 2009, ApJ, 694, 396
90
Brooks, A. M. et al. 2011, ApJ, 728, 51
Brooks, A. M., & Zolotov, A. 2012, ArXiv e-prints
Bruce, V. A. et al. 2012, MNRAS, 427, 1666
Buat, V. 1992, A&A, 264, 444
Buat, V., Donas, J., & Deharveng, J. M. 1987, A&A, 185, 33
Bullock, J. S., Kolatt, T. S., Sigad, Y., Somerville, R. S., Kravtsov, A. V., Klypin, A. A.,
Primack, J. R., & Dekel, A. 2001, MNRAS, 321, 559
Bundy, K., Ellis, R., & Conselice, C. 2005, astro–ph/0502204
Burkert, A. 1995, ApJ, 447, L25
Cannon, J. M. et al. 2011, ApJ, 739, L22
Cappellari, M. et al. 2012, Nature, 484, 485
Ceverino, D., & Klypin, A. 2009, ApJ, 695, 292
Chabrier, G. 2003, PASP, 115, 763
Choi, J.-H., & Nagamine, K. 2011, MNRAS, 410, 2579
Christensen, C., Governato, F., Quinn, T., Brooks, A. M., Fisher, D. B., Shen, S., McCleary,
J., & Wadsley, J. 2012a, ArXiv e-prints
—. 2012b, ArXiv e-prints
Christensen, C., Quinn, T., Governato, F., Stilp, A., Shen, S., & Wadsley, J. 2012c, MNRAS, 425, 3058
—. 2012d, MNRAS, 425, 3058
Cloet-Osselaer, A., De Rijcke, S., Schroyen, J., & Dury, V. 2012, MNRAS, 2952
91
Colı́n, P., Avila-Reese, V., Vázquez-Semadeni, E., Valenzuela, O., & Ceverino, D. 2010,
ApJ, 713, 535
Conroy, C., & van Dokkum, P. G. 2012, ApJ, 760, 71
Conroy, C., Wechsler, R. H., & Kravtsov, A. V. 2006, ApJ, 647, 201
Croton, D. J. 2009, MNRAS, 394, 1109
Dalcanton, J. J. 1998, ApJ, 495, 251
Davis, M., Efstathiou, G., Frenk, C. S., & White, S. D. M. 1985, ApJ, 292, 371
de Souza, R. S., Rodrigues, L. F. S., Ishida, E. E. O., & Opher, R. 2011, MNRAS, 415,
2969
Dekel, A. et al. 2009, Nature, 457, 451
Dekel, A., & Silk, J. 1986, ApJ, 303, 39
Di Matteo, T., Springel, V., & Hernquist, L. 2005, Nature, 433, 604
Dutton, A. A., Macciò, A. V., Mendel, J. T., & Simard, L. 2013, MNRAS, 432, 2496
Eke, V. R., Navarro, J. F., & Steinmetz, M. 2001, ApJ, 554, 114
Fakhouri, O., Ma, C.-P., & Boylan-Kolchin, M. 2010, MNRAS, 406, 2267
Fall, S. M., & Efstathiou, G. 1980, MNRAS, 193, 189
Ferland, G. J., Korista, K. T., Verner, D. A., Ferguson, J. W., Kingdon, J. B., & Verner,
E. M. 1998a, PASP, 110, 761
—. 1998b, PASP, 110, 761
Frenk, C. S., White, S. D. M., Efstathiou, G., & Davis, M. 1985, Nature, 317, 595
Geha, M., Blanton, M. R., Masjedi, M., & West, A. A. 2006, ApJ, 653, 240
92
Geller, M. J., Diaferio, A., Kurtz, M. J., Dell’Antonio, I. P., & Fabricant, D. G. 2012, AJ,
143, 102
Genel, S., Vogelsberger, M., Nelson, D., Sijacki, D., Springel, V., & Hernquist, L. 2013,
MNRAS
Genzel, R. et al. 2012, ApJ, 746, 69
Gill, S. P. D., Knebe, A., & Gibson, B. K. 2004, MNRAS, 351, 399
Gilmore, G., Hernandez, X., & Valls-Gabaud, D. 2000, Ap&SS, 272, 61
Giovanelli, R. et al. 2005, AJ, 130, 2598
Gnedin, N. Y. 2000, ApJ, 542, 535
Gnedin, N. Y., Tassis, K., & Kravtsov, A. V. 2009, ApJ, 697, 55
Governato, F. et al. 2010, Nature, 463, 203
—. 2009, MNRAS, 398, 312
Governato, F., Willman, B., Mayer, L., Brooks, A., Stinson, G., Valenzuela, O., Wadsley,
J., & Quinn, T. 2007, MNRAS, 374, 1479
Governato, F. et al. 2012, MNRAS, 422, 1231
Graham, A. W., Driver, S. P., Petrosian, V., Conselice, C. J., Bershady, M. A., Crawford,
S. M., & Goto, T. 2005, AJ, 130, 1535
Grogin, N. A. et al. 2011, ApJS, 197, 35
Guedes, J., Callegari, S., Madau, P., & Mayer, L. 2011, ApJ, 742, 76
Guo, Q., White, S., Li, C., & Boylan-Kolchin, M. 2010, MNRAS, 404, 1111
Haardt, F., & Madau, P. 1996, ApJ, 461, 20
93
Heavens, A., Panter, B., Jimenez, R., & Dunlop, J. 2004, Nature, 428, 625
Hennebelle, P., & Chabrier, G. 2008, ApJ, 684, 395
Hopkins, A. M., McClure-Griffiths, N. M., & Gaensler, B. M. 2008, ApJ, 682, L13
Hopkins, P. F. 2012a, MNRAS, 423, 2016
—. 2012b, MNRAS, 423, 2037
—. 2013, MNRAS, 433, 170
Hopkins, P. F., Hernquist, L., Cox, T. J., Di Matteo, T., Martini, P., Robertson, B., &
Springel, V. 2005, ApJ, 630, 705
Hopkins, P. F., Keres, D., Murray, N., Quataert, E., & Hernquist, L. 2011a, ArXiv e-prints
Hopkins, P. F., Quataert, E., & Murray, N. 2011b, ArXiv e-prints
Hopkins, P. F. et al. 2009, MNRAS, 397, 802
Huang, S., Haynes, M. P., Giovanelli, R., Brinchmann, J., Stierwalt, S., & Neff, S. G. 2012,
ArXiv e-prints
Jenkins, A. et al. 1998, ApJ, 499, 20
Johansson, P. H., Naab, T., & Ostriker, J. P. 2012, ApJ, 754, 115
Jonsson, P. 2006, MNRAS, 372, 2
Jonsson, P., Groves, B. A., & Cox, T. J. 2010, MNRAS, 403, 17
Katz, N. 1992, ApJ, 391, 502
Katz, N., & White, S. D. M. 1993, ApJ, 412, 455
Kauffmann, G. et al. 2003, MNRAS, 341, 33
Kaufmann, D. E. 1993, PhD thesis, Florida University
94
Keres, D., Vogelsberger, M., Sijacki, D., Springel, V., & Hernquist, L. 2011, ArXiv e-prints
Kereš, D., Katz, N., Davé, R., Fardal, M., & Weinberg, D. H. 2009, MNRAS, 396, 2332
Klypin, A., Kravtsov, A. V., Valenzuela, O., & Prada, F. 1999, ApJ, 522, 82
Knollmann, S. R., & Knebe, A. 2009, ApJS, 182, 608
Koekemoer, A. M. et al. 2011, ApJS, 197, 36
Kormendy, J., & Sanders, D. B. 1992, ApJ, 390, L53
Kroupa, P., & Weidner, C. 2003, ApJ, 598, 1076
Krumholz, M. R. 2011, ApJ, 743, 110
Krumholz, M. R., McKee, C. F., & Tumlinson, J. 2009, ApJ, 699, 850
Kuhlen, M., Diemand, J., Madau, P., & Zemp, M. 2008, Journal of Physics Conference
Series, 125, 012008
Kuhlen, M., Krumholz, M., Madau, P., Smith, B., & Wise, J. 2011, ArXiv e-prints
Lada, C. J., & Lada, E. A. 2003, ARA&A, 41, 57
Larson, R. B. 2005, MNRAS, 359, 211
Leauthaud, A. et al. 2011, ArXiv e-prints
—. 2012, ApJ, 744, 159
Leitherer, C. et al. 1999, ApJS, 123, 3
Leitner, S. N. 2012, ApJ, 745, 149
Macciò, A. V., Stinson, G., Brook, C. B., Wadsley, J., Couchman, H. M. P., Shen, S.,
Gibson, B. K., & Quinn, T. 2012, ApJ, 744, L9
Madau, P., Diemand, J., & Kuhlen, M. 2008, ApJ, 679, 1260
95
Mandelbaum, R., Seljak, U., Kauffmann, G., Hirata, C. M., & Brinkmann, J. 2006, MNRAS,
368, 715
Maraston, C. et al. 2012, ArXiv e-prints
Martin, C. L. 1999, ApJ, 513, 156
Mashchenko, S., Couchman, H. M. P., & Wadsley, J. 2006, Nature, 442, 539
Mateo, M. L. 1998, ARA&A, 36, 435
McCarthy, I. G., Schaye, J., Bower, R. G., Ponman, T. J., Booth, C. M., Dalla Vecchia, C.,
& Springel, V. 2011, MNRAS, 412, 1965
McCarthy, I. G., Schaye, J., Font, A. S., Theuns, T., Frenk, C. S., Crain, R. A., & Dalla
Vecchia, C. 2012, ArXiv e-prints
Mo, H. J., Mao, S., & White, S. D. M. 1998, MNRAS, 295, 319
Moore, B. 1994, Nature, 370, 629
Moore, B., Governato, F., Quinn, T., Stadel, J., & Lake, G. 1998, ApJ, 499, L5+
Moore, B., Quinn, T., Governato, F., Stadel, J., & Lake, G. 1999a, MNRAS, 310, 1147
—. 1999b, MNRAS, 310, 1147
More, S., van den Bosch, F. C., & Cacciato, M. 2009, MNRAS, 392, 917
Moster, B. P., Naab, T., & White, S. D. M. 2012, ArXiv e-prints
Moster, B. P., Somerville, R. S., Maulbetsch, C., van den Bosch, F. C., Macciò, A. V.,
Naab, T., & Oser, L. 2010, ApJ, 710, 903
Munshi, F., Christensen, C., Quinn, T. R., Governato, F., Wadsley, J., Loebman, S., &
Shen, S. 2013a, ArXiv e-prints
Munshi, F. et al. 2013b, ApJ, 766, 56
96
Narayanan, D., & Davé, R. 2012, MNRAS, 423, 3601
Narayanan, D., & Hopkins, P. F. 2013, MNRAS, 433, 1223
Nickerson, S., Stinson, G., Couchman, H. M. P., Bailin, J., & Wadsley, J. 2011, MNRAS,
415, 257
Ogiya, G., & Mori, M. 2012, ArXiv e-prints
Oh, S.-H., Brook, C., Governato, F., Brinks, E., Mayer, L., de Blok, W. J. G., Brooks, A.,
& Walter, F. 2011a, AJ, 142, 24
—. 2011b, AJ, 142, 24
Okamoto, T., Gao, L., & Theuns, T. 2008, MNRAS, 390, 920
Oppenheimer, B. D., Davé, R., Kereš, D., Fardal, M., Katz, N., Kollmeier, J. A., & Weinberg, D. H. 2010, MNRAS, 406, 2325
Oser, L., Naab, T., Ostriker, J. P., & Johansson, P. H. 2012, ApJ, 744, 63
Padoan, P., Haugbølle, T., & Nordlund, Å. 2012, ApJ, 759, L27
Padoan, P., Nordlund, A., & Jones, B. J. T. 1997, MNRAS, 288, 145
Panter, B., Heavens, A. F., & Jimenez, R. 2004, MNRAS, 355, 764
Pasetto, S., Grebel, E. K., Berczik, P., Spurzem, R., & Dehnen, W. 2010, A&A, 514, A47
Pforr, J., Maraston, C., & Tonini, C. 2012, ArXiv e-prints
Piontek, F., & Steinmetz, M. 2011, MNRAS, 410, 2625
Pontzen, A., & Governato, F. 2012a, ArXiv e-prints
—. 2012b, MNRAS, 421, 3464
Pontzen, A. et al. 2008, MNRAS, 390, 1349
97
Quinn, T., Katz, N., & Efstathiou, G. 1996, MNRAS, 278, L49
Reddick, R. M., Wechsler, R. H., Tinker, J. L., & Behroozi, P. S. 2012, ArXiv e-prints
Reed, D., Gardner, J., Quinn, T., Stadel, J., Fardal, M., Lake, G., & Governato, F. 2003,
MNRAS, 346, 565
Reed, D. S., Governato, F., Quinn, T., Stadel, J., & Lake, G. 2007, MNRAS, 378, 777
Reid, B. A. et al. 2010, MNRAS, 404, 60
Robertson, B., Yoshida, N., Springel, V., & Hernquist, L. 2004, ApJ, 606, 32
Sales, L. V., Navarro, J. F., Theuns, T., Schaye, J., White, S. D. M., Frenk, C. S., Crain,
R. A., & Dalla Vecchia, C. 2012, MNRAS, 3041
Salim, S. et al. 2007, ApJS, 173, 267
Santini, P. et al. 2012, A&A, 538, A33
Sawala, T., Frenk, C. S., Crain, R. A., Jenkins, A., Schaye, J., Theuns, T., & Zavala, J.
2012, ArXiv e-prints
Sawala, T., Guo, Q., Scannapieco, C., Jenkins, A., & White, S. 2011, MNRAS, 413, 659
Scannapieco, C., Gadotti, D. A., Jonsson, P., & White, S. D. M. 2010, MNRAS, 407, L41
Scannapieco, C. et al. 2012, MNRAS, 423, 1726
Scannapieco, C., White, S. D. M., Springel, V., & Tissera, P. B. 2011, MNRAS, 417, 154
Schawinski, K., Treister, E., Urry, C. M., Cardamone, C. N., Simmons, B., & Yi, S. K.
2011, ApJ, 727, L31+
Shen, S., Wadsley, J., & Stinson, G. 2010, MNRAS, 407, 1581
Sheth, R. K., Mo, H. J., & Tormen, G. 2001, MNRAS, 323, 1
98
Shimasaku, K. et al. 2001, AJ, 122, 1238
Sijacki, D., Springel, V., Di Matteo, T., & Hernquist, L. 2007, MNRAS, 380, 877
Somerville, R. S., Hopkins, P. F., Cox, T. J., Robertson, B. E., & Hernquist, L. 2008,
MNRAS, 391, 481
Somerville, R. S., & Primack, J. R. 1999, MNRAS, 310, 1087
Springel, V., Di Matteo, T., & Hernquist, L. 2005a, MNRAS, 361, 776
Springel, V., & Hernquist, L. 2003, MNRAS, 339, 312
Springel, V. et al. 2008, Nature, 456, 73
—. 2005b, Nature, 435, 629
Stinson, G., Brook, C., Macciò, A. V., Wadsley, J., Quinn, T. R., & Couchman, H. M. P.
2012, ArXiv e-prints
Stinson, G., Seth, A., Katz, N., Wadsley, J., Governato, F., & Quinn, T. 2006, MNRAS,
373, 1074
Teyssier, R. et al. 2009, A&A, 497, 335
Teyssier, R., Pontzen, A., Dubois, Y., & Read, J. 2012, ArXiv e-prints
Thacker, R. J., & Couchman, H. M. P. 2000, ApJ, 545, 728
Toomre, A., & Toomre, J. 1972, ApJ, 178, 623
Treu, T., Auger, M. W., Koopmans, L. V. E., Gavazzi, R., Marshall, P. J., & Bolton, A. S.
2010, ApJ, 709, 1195
Trujillo-Gomez, S., Klypin, A., Primack, J., & Romanowsky, A. J. 2011, ApJ, 742, 16
Vale, A., & Ostriker, J. P. 2004, MNRAS, 353, 189
99
van Daalen, M. P., Angulo, R. E., & White, S. D. M. 2012, ArXiv e-prints
van Dokkum, P. G. et al. 2011, ApJ, 743, L15
van Dokkum, P. G., & Conroy, C. 2011, ApJ, 735, L13
Vázquez, G. A., & Leitherer, C. 2005, ApJ, 621, 695
Wadsley, J. W., Stadel, J., & Quinn, T. 2004, New Astronomy, 9, 137
Wang, J., & White, S. D. M. 2009, MNRAS, 396, 709
Wang, P., Li, Z., Abel, T., & Nakamura, F. 2010, ApJ, 709, 27
Warren, B. E., Jerjen, H., & Koribalski, B. S. 2006a, AJ, 131, 2056
—. 2006b, AJ, 131, 2056
Weidner, C., Ferreras, I., Vazdekis, A., & La Barbera, F. 2013, MNRAS, 435, 2274
Weisz, D. R. et al. 2012, ApJ, 744, 44
White, S. D. M., Davis, M., Efstathiou, G., & Frenk, C. S. 1987, Nature, 330, 451
White, S. D. M., & Frenk, C. S. 1991, ApJ, 379, 52
White, S. D. M., & Rees, M. J. 1978, MNRAS, 183, 341
Willman, B. et al. 2005, ApJ, 626, L85
Willman, B., Governato, F., Dalcanton, J. J., Reed, D., & Quinn, T. 2004, MNRAS, 353,
639
Wise, J. H., Abel, T., Turk, M. J., Norman, M. L., & Smith, B. D. 2012a, ArXiv e-prints
Wise, J. H., Turk, M. J., Norman, M. L., & Abel, T. 2012b, ApJ, 745, 50
Yang, X., Mo, H. J., van den Bosch, F. C., Zhang, Y., & Han, J. 2012, ApJ, 752, 41
100
Yasuda, N. et al. 2001, AJ, 122, 1104
Zheng, Z., Coil, A. L., & Zehavi, I. 2007, ApJ, 667, 760
Zolotov, A. et al. 2012, ApJ, 761, 71
101
VITA
Ferah Munshi began her study of Astronomy after her third year at the University of
California, Berkeley under the supervision of Professor Saul Permlmutter and Dr. Andrew
West. After graduating with a Bachelor of Arts from Berkeley in 2008, she made her way to
the University of Washington to begin the graduate program in the Astronomy Department.
She completed her doctorate at the University of Washington in December 2013.